Re: Apache http / mod_rewrite / mod_jk
Yoav, Do you know if this is supposed to work? JkMount /myapp/*;jsessionid=* workerX It has allegedly worked for some other people. I don't know if it's supposed to work. Is it difficult to test? I have tested it. It does not work for me. Others claim that it does work. I'm wondering if I have something wrong. I've been using mod_jk for years with no problems. I have multiple workers and multiple applications, all running fine. This particular directive seems to be ignored on my setup. I guess I'll just look at the module source and figure it out myself. -chris signature.asc Description: OpenPGP digital signature
Re: Apache http / mod_rewrite / mod_jk
Kenneth, Instead of directly redirect to mod_jk, can you just use mod_write to insert a directory prefix which can trigger mod_jk? That's an idea. If I did that, how could I get Tomcat to figure out the right path, then? Can I do a reverse rewrite in Tomcat? Thanks, -chris signature.asc Description: OpenPGP digital signature
Re: Apache http / mod_rewrite / mod_jk
All,
JkMount /myapp/*;jsessionid=* workerX
This particular directive seems to be ignored on my setup. I guess I'll
just look at the module source and figure it out myself.
It turns out that mod_jk does not consider ;jsessionid= part of the
URL when it does matching. :(
From mod_jk.log:
[jk_uri_worker_map.c (486)]: Into jk_uri_worker_map_t::map_uri_to_worker
[jk_uri_worker_map.c (500)]: Attempting to map URI
'/app/includes/css/styles.css'
[jk_uri_worker_map.c (618)]: jk_uri_worker_map_t::map_uri_to_worker,
done without a match
[jk_uri_worker_map.c (486)]: Into jk_uri_worker_map_t::map_uri_to_worker
[jk_uri_worker_map.c (500)]: Attempting to map URI '/app/images/select.gif'
[jk_uri_worker_map.c (618)]: jk_uri_worker_map_t::map_uri_to_worker,
done without a match
Both of these URLs definately have the session id attached to them.
jk_uri_worker_map.c::map_uri_to_worker contains the following code
torards the beginning of the function:
char *url_rewrite = strstr(uri, JK_PATH_SESSION_IDENTIFIER);
if(url_rewrite) {
*url_rewrite = '\0';
}
jk_no2slash(uri);
So, they completely strip-out the sessionid before checking
(JK_PATH_SESSION_IDENTIFIER = ;jsessionid).
My conclusion is that there's no support for this kind of think in
mod_jk. The following matches are supported (from the code docs):
* Exact Context - /exact/uri=worker e.g. /examples/do*=ajp12 *
* Context Based - /context/*=worker e.g. /examples/*=ajp12 *
* Context and suffix -/context/*.suffix=worker e.g. /examples/*.jsp=ajp12*
It's not a simple fix to add the kind of checking I'm talking about.
I'll poke around to find out how this might be done.
If there are any mod_jk hackers on the list reading this thread, please
let me know. Also, if anyone knows if this works in other connectors
(i.e. jk2), please let me know.
I'm putting all of this in a post so that it gets into the archives.
I'm also adding this sentence to help out with searches: The problem is
with jsessionid, Apache, mod_jk and is most often characterized by
broken images and failure to load stylesheets when URLs are encoded in
Java doe these resources.
Thanks,
-chris
signature.asc
Description: OpenPGP digital signature
Re: Apache http / mod_rewrite / mod_jk [hacked SOLUTION]
All,
Here is my solution to the jsessionid issue discussed in this thread.
The problem is that when Tomcat encodes URLs without knowing if the
browser supports cookies, it adds ;jsessionid=BLAHBLAH to each encoded
URL. This is perfectly normal behavior.
However, when using Apache httpd, Apache does not properly handle the
addition of the ;jsessionid information, and fails to serve up the
proper file.
One of the simplest solutions is to simply pass requests including the
jsessionid information to Tomcat and let Tomcat handle them.
mod_jk does not support this type of behavior, since it only does simple
context matches (like /context/*, suffix-based context matches
(/context/*.suffix) or exact matches (/context/j_security_check).
I have modified mod_jk to allow URIs of this special form:
/context/*;jsessionid*
So you can actually do this in your configuration file:
JkMount /context/*;jsessionid* my_worker
The syntax matches what I figured was the most natural way to write this
in a configuration file. Note that this fix does not actually let you
use wildcards as flexibly as the config directive would lead you to
believe. It only works with the specific string *;jsessionid* after
the context and slash character.
Below is the diff for jk_uri_worker_map.c. It works on my configuration.
I hope it may help others. I would love to receive any constructive
criticism of this patch. Feel free to email me directly.
All the best,
-chris
=
[EMAIL PROTECTED] common]$ diff -bc jk_uri_worker_map.c jk_uri_worker_map.c.new
*** jk_uri_worker_map.c 2003-09-06 11:37:21.0 -0400
--- jk_uri_worker_map.c.new 2004-04-08 10:07:36.0 -0400
***
*** 80,85
--- 80,86
#define MATCH_TYPE_GENERAL_SUFFIX (3) /* match all URIs of the form
*ext */
/* match all context path URIs with a path component suffix */
#define MATCH_TYPE_CONTEXT_PATH (4)
+ #define MATCH_TYPE_JSESSIONID (5) /* match all URLs that contains a
jsessionid */
struct uri_worker_record {
/* Original uri for logging */
***
*** 300,305
--- 301,316
Into
jk_uri_worker_map_t::uri_worker_map_open,
suffix rule %s.%s=%s was added\n,
uri, asterisk + 3, worker);
+ } else if(0 == strncmp(/* JK_PATH_SESSION_IDENTIFIER
*,asterisk,14) ) {
+ /* jsessionid interceptor */
+ asterisk[1] = '\0';
+ uwr-worker_name = worker;
+ uwr-context = uri;
+ uwr-match_type = MATCH_TYPE_JSESSIONID;
+ jk_log(l, JK_LOG_DEBUG,
+ Into jk_uri_worker_map_t::uri_worker_map_open,
+ jsessionid rule %s*;jsessionid*=%s was added\n,
+ uri, worker);
} else if ('\0' != asterisk[2]) {
/* general suffix rule */
asterisk[1] = '\0';
***
*** 478,483
--- 489,522
*d = '\0';
}
+ uri_worker_record_t *check_jsessionid_matches(char *uri,
+ jk_uri_worker_map_t *uw_map,
+ jk_logger_t *l)
+ {
+ unsigned i;
+ uri_worker_record_t *match = NULL;
+
+ for(i = 0 ; i uw_map-size ; i++) {
+ uri_worker_record_t *uwr = uw_map-maps[i];
+
+ if(MATCH_TYPE_JSESSIONID == uwr-match_type) {
+ /* Check context match */
+
+ if(0 == strncmp(uwr-context,
+ uri,
+ uwr-ctxt_len)) {
+ /* Looks like a match: URI contains ;jsessionid, context
matches */
+
+ jk_log(l, JK_LOG_DEBUG, check_jsessionid_matches: Found context
match: %s\n, uwr-context);
+
+ match = uwr;
+ break;
+ }
+ } /* MATCH_TYPE_JSESSIONID */
+ } /* foreach mapping */
+
+ return match;
+ }
char *map_uri_to_worker(jk_uri_worker_map_t *uw_map,
char *uri,
***
*** 493,498
--- 532,544
char *url_rewrite = strstr(uri, JK_PATH_SESSION_IDENTIFIER);
if(url_rewrite) {
+ uri_worker_record_t *match = check_jsessionid_matches(uri,
uw_map, l);
+ if(NULL != match) {
+ jk_log(l, JK_LOG_DEBUG, Mapped URI %s to context %s using
jsessionid match\n, uri, match-context);
+
+ return match-worker_name;
+ }
+
*url_rewrite = '\0';
}
jk_no2slash(uri);
signature.asc
Description: OpenPGP digital signature
Re: Apache http / mod_rewrite / mod_jk [hacked SOLUTION]
All,
There was a bug in the patch that I posted. Below is the final patch I
will post to the group.
-chris
*** common/jk_uri_worker_map.c 2003-09-06 11:37:21.0 -0400
--- common/jk_uri_worker_map.c.new 2004-04-08 14:23:10.0 -0400
***
*** 80,85
--- 80,86
#define MATCH_TYPE_GENERAL_SUFFIX (3) /* match all URIs of the form
*ext */
/* match all context path URIs with a path component suffix */
#define MATCH_TYPE_CONTEXT_PATH (4)
+ #define MATCH_TYPE_JSESSIONID (5)
struct uri_worker_record {
/* Original uri for logging */
***
*** 300,305
--- 301,317
Into
jk_uri_worker_map_t::uri_worker_map_open,
suffix rule %s.%s=%s was added\n,
uri, asterisk + 3, worker);
+ } else if(0 == strncmp(/* JK_PATH_SESSION_IDENTIFIER
*,asterisk,14) ) {
+ /* jsessionid rule */
+ asterisk[1] = '\0';
+ uwr-worker_name = worker;
+ uwr-context = uri;
+ uwr-suffix = \0; /* avoids some problems */
+ uwr-match_type = MATCH_TYPE_JSESSIONID;
+ jk_log(l, JK_LOG_DEBUG,
+ Into jk_uri_worker_map_t::uri_worker_map_open,
+ jsessionid rule %s*;jsessionid*=%s was added\n,
+ uri, worker);
} else if ('\0' != asterisk[2]) {
/* general suffix rule */
asterisk[1] = '\0';
***
*** 478,483
--- 490,523
*d = '\0';
}
+ uri_worker_record_t *check_jsessionid_matches(char *uri,
+ jk_uri_worker_map_t *uw_map,
+ jk_logger_t *l)
+ {
+ unsigned i;
+ uri_worker_record_t *match = NULL;
+
+ for(i = 0 ; i uw_map-size ; i++) {
+ uri_worker_record_t *uwr = uw_map-maps[i];
+
+ if(MATCH_TYPE_JSESSIONID == uwr-match_type) {
+ /* Check context match */
+
+ if(0 == strncmp(uwr-context,
+ uri,
+ uwr-ctxt_len)) {
+ /* Looks like a match: URI contains ;jsessionid, context matches */
+
+ jk_log(l, JK_LOG_DEBUG, check_jsessionid_matches: Found context
match: %s\n, uwr-context);
+
+ match = uwr;
+ break;
+ }
+ } /* MATCH_TYPE_JSESSIONID */
+ } /* foreach mapping */
+
+ return match;
+ }
char *map_uri_to_worker(jk_uri_worker_map_t *uw_map,
char *uri,
***
*** 493,498
--- 533,545
char *url_rewrite = strstr(uri, JK_PATH_SESSION_IDENTIFIER);
if(url_rewrite) {
+ uri_worker_record_t *match = check_jsessionid_matches(uri,
uw_map, l);
+ if(NULL != match) {
+ jk_log(l, JK_LOG_DEBUG, Mapped URI %s to context %s using
jsessioni match\n, uri, match-context);
+
+ return match-worker_name;
+ }
+
*url_rewrite = '\0';
}
jk_no2slash(uri);
***
*** 559,564
--- 606,615
}
}
}
+ } else if (MATCH_TYPE_JSESSIONID == uwr-match_type) {
+
+ /* do nothing -- these have already been handled */
+
} else /* suffix match */ {
int suffix_start;
signature.asc
Description: OpenPGP digital signature
Apache http / mod_rewrite / mod_jk
All, The archives show this questions being asked all the time, but with no useful responses. Please let me know if this is a known unresolved or unresolvable issue. All solutions posted anywhere for jsessionid makes Apache go beaindead apparently use a mod_rewrite incantation similar to the following: IfModule mod_rewrite.c RewriteEngine on # Force URLs with a jsessionid to go to Tomcat. Necessary because # Apache doesn't recognise that the semi-colon is special. RewriteRule ^(/.*;jsessionid=.*)$ $1 [T=jserv-servlet] /IfModule While I'm sure this worked out great the people using mod_jserv back in 1997, it does not work for mod_jk. For one thing, it does not even let you specify which worker to use :( Back in the day, Craig responded by pointing to a Tomcat FAQ entry which no longer exists, but presumably had something to do with Apache's mod_rewrite. On the other hand, a solution was posted (and confirmed by some readers) that the following works: JkMount /test/*;jsessionid=* ajp13 This seems very obvious, and there's a caveat about how it might not work on older versions of mod_jk. It apparently does not work for me. I'm using mod_jk (not mod_jk2), version 1.2.5 (current release) on Apache 2.0.48 as a dynamic module on Linux -- everything compiled myself with nothing out of the ordinary. Can anyone offer any advice? I've just been sucking it up and ignoring this problem for a while, now (years). Is there actually a solution out there for this? Am I just mistyping the JkMount configuration? Anyone, please help. Thanks, -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Apache http / mod_rewrite / mod_jk
Yoav, I have no clue as to your actual question, but I'm curious: The archives show this questions being asked all the time, but with no How do you define all the time in the statement above? Like this: Dubious responses: http://www.mail-archive.com/[EMAIL PROTECTED]/msg84808.html http://www.mail-archive.com/[EMAIL PROTECTED]/msg74207.html http://www.mail-archive.com/[EMAIL PROTECTED]/msg84936.html (references a URL @ Google Groups with a simple solution that does not appear to work for me) References Craig's pointers: http://www.mail-archive.com/[EMAIL PROTECTED]/msg78826.html No responses: http://www.mail-archive.com/[EMAIL PROTECTED]/msg75615.html That's just in the tomcat-user archives. Similar discussions occur in other forums (fora?), like jGuru. Do you know if this is supposed to work? JkMount /myapp/*;jsessionid=* workerX It has allegedly worked for some other people. -chris signature.asc Description: OpenPGP digital signature
Load testing using ApacheBench
All, I've started load testing my application using ApacheBench and I think I must be missing something. I only heard about this tool very recently, so I have pretty much no experience with it. I was getting some odd behavior (i.e. very long response times compared to what I observe when just hitting the app myself via a browser) so I decided to lower the number of requests. The response times became very erratic. I tried 10 requests. Sometimes, I'd get all of them back within a few (10) ms. Other times, it would take 5-10 seconds for the last few requests. I figured I'd profile it. So I did. But I continued to get strange results. For example, making a single request would take 30 seconds and create a ton of objects, etc. During the 30-second test, I could use my browser and make the same request with a reasonable response time (5 sec). Bizarre. So, I started studying the output of ab: (Sorry for the full output, but I need some advice and I figured that more info would be better). === localhost[root]:/usr/local/src/httpd-2.0.48/support$ ./ab -n 1 http://host/app/simple_search.do?query=testresourceType=-1search=Search; This is ApacheBench, Version 2.0.40-dev $Revision: 1.121.2.4 $ apache-2.0 Copyright (c) 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/ Copyright (c) 1998-2002 The Apache Software Foundation, http://www.apache.org/ Benchmarking host (be patient).done Server Software:Apache/2.0.46 Server Hostname:host Server Port:8180 Document Path: /app/simple_search.do?query=testresourceType=-1search=Search Document Length:14140 bytes Concurrency Level: 1 Time taken for tests: 108.739776 seconds Complete requests: 1 Failed requests:215 (Connect: 0, Length: 0, Exceptions: 215) Write errors: 0 Total transferred: 3069721 bytes HTML transferred: 3017449 bytes Requests per second:0.01 [#/sec] (mean) Time per request: 108739.777 [ms] (mean) Time per request: 108739.777 [ms] (mean, across all concurrent requests) Transfer rate: 27.56 [Kbytes/sec] received Connection Times (ms) min mean[+/-sd] median max Connect:00 0.0 0 0 Processing: 1422 1422 0.0 14221422 Waiting: 1059 1059 0.0 10591059 Total: 1422 1422 0.0 14221422 localhost[root]:/usr/local/src/httpd-2.0.48/support$ === WTF? 215 failed requests? I only asked for one! Not only that, but I tailed my apache request log during a subsequent identical test and found lots of these: 192.168.1.43 - - [03/Apr/2004:13:29:49 -0500] GET /app/simple_search.do?query=testresourceType=-1search=Search HTTP/1.0 200 14146 192.168.1.43 - - [03/Apr/2004:13:29:49 -0500] GET /app/simple_search.do?query=testresourceType=-1search=Search HTTP/1.0 200 14140 192.168.1.43 - - [03/Apr/2004:13:29:50 -0500] GET /app/simple_search.do?query=testresourceType=-1search=Search HTTP/1.0 200 14140 192.168.1.43 - - [03/Apr/2004:13:29:50 -0500] GET /app/simple_search.do?query=testresourceType=-1search=Search HTTP/1.0 200 14140 WTF? Response=200 looks good to me! Why am I getting all of these failed requests? Am I totally missing something? This tool seems stupidly simple to use. Am I not reading the output correctly? I tried to use the -v switch for verbose troubleshooting information, but I don't know what the following argument should be. I tried everything I could think of like numbers and stuff like debug error, etc. -- no dice. Please, can someone help me understand what I'm doing wrong? Thanks a /ton/ in advance, -chris signature.asc Description: OpenPGP digital signature
ChannelSocket.read -- always runnable?
All, I'm currently profiling my app running on: RH Linux 9 / kernel 2.4.20-8 Sun JDK 1.4.2 Tomcat 4.1.29 I'm using Borland's OptimizeIt v4.2 (a bit old, but hey! it's expensive!). Technically, I'm running the host on my server (on Linux) and the OptimizeIt profiling client on a separate windoze box and attaching to the process over the network. It's really irrelevant, but often people ask about these things. When I do a CPU profile (really a what are threads doing profile) during a test where I just make requests as fast as possible, one at a time, I'm seeing that most of the time spent by runnable (RequestProcessor) threads is in this method: java.net.SocketInputStream.read() which gets called by org.apache.jk.common.ChannelSocket.read. I assume that this is just the Java part of the JK connector sitting there waiting for a connection from Apache. My question is this: why is it runnable? A VM thread dump says that they're *all* runnable, instead of sleeping, waiting for data. Is this just the way that Java does its sockets? I don't know a ton about the Java socket and I/O implementations, but I didn't think that it did a spin-wait on stream reading (rather than a blocking read, which would... well, block). Anyhow, it makes the thread profile look very weird, because all of the threads are runnable all of the time. Makes it tough to find out which threads are actually doing any work :( It this behavior consistent across Java versions? Are different socket implementations used with different connectors? It would be nice if fewer threads were runnable when they're really not doing anything. If someone can give me a good reason for these threads being runnable, I'd love to hear it. Thanks, -chris signature.asc Description: OpenPGP digital signature
Re: Native libraries configuration under tomcat.
Nick,
JMagick comes as a semi-stub library .jar archive and a native library
libJMagick.so - under the command line my code works (following ldconfig),
but I get errors in the catalina.out complaining about being unable to find
the library (details at the end of this message). the key clause being:
Caused by: java.lang.RuntimeException: Can't load MagickLoader (class not
found)
at magick.Magick.clinit(Magick.java:25)
Essentially, having trawled through the various documentation, I haven't
worked out how to direct Tomcat to find/use the .so library, so any pointers
to how to do this would be great.
I'm not sure where it is, but I'm sure that Tomcat has a directory
that's specified as a search path for shared libs. Your JMagick class's
clinit (public static { }) method is bombing probably because a call
to System.loadLibrary() can't find the shared lib.
Can anyone say if Tomcat itself provides 'another' directory where
shared objects can go?
You can put it in the SDK somewhere, but that's ugly.
Finally, let me issue a word of warning: using native code from a webapp
is pretty dangerous: if the native code, say, segfaults, then Tomcat
will go down along with all your webapps. Also, you have to be careful:
the JVM gets very upset if you call System.loadLibrary() more than once
for a given library. So, you should take care to have your JMagick
library sitting in a common area, rather than your webapp's WEB-INF/lib
directory. (Unless you promose never to re-start your webapp without
re-starting Tomcat ;)
-chris
signature.asc
Description: OpenPGP digital signature
Re: apache 1.3 tomcat 5.0.19
Emerson, a second question: may I use the same mod_jk.so I used with tomcat 3.2.3?? The configuration of mod_jk is independent of the Tomcat version. As long as the version of Tomcat that you are using supports mod_jk (really ajp13), you can use mod_jk with it (including any existing configuration). apj13 is supported by pretty much every version of Tomcat that you can get your hands on. -chris signature.asc Description: OpenPGP digital signature
Re: JNDI Datasource Reference in DD Not Necessary?
Kawthar, In my web.xml: ... res-ref-namejdbc/mySQLDatabase/res-ref-name Does this looks correct? From the exception message, which class name is it looking for? It looks like I have the same stuff that you do. This is my entire resource-ref element in web.xml: resource-ref descriptionDataSource for [the] application/description res-ref-namejdbc/[ref-name]/res-ref-name res-typejavax.sql.DataSource/res-type res-authContainer/res-auth /resource-ref (Note that I'm using Tomcat 4.1.30) -chris signature.asc Description: OpenPGP digital signature
Re: Why Tomcat (4.1.29) cannot find third party package's classes
Ari, is there any particular reason why all classes in third party jar-package (Exolab Castor in my case) can not be imported in my JSP pages on new Tomcat (4.1.29 and Java build 1.4.2_03-b02)? It works fine with the old version (Tomcat 4.0.1 and Java build 1.3.1_02-b02). Where is your Castor JAR file located? -chris signature.asc Description: OpenPGP digital signature
Re: Filters and load-on-startup
Raphael, The webapp has a login filter that needs a connection from the connection pool. The webapp loads the login filter first. I don't really want to remove the filter because its integral to the application's design. Is there another solution, to getting the connection pool instantiated before the filter? Is this really a problem? Even though the filter needs a DB connection to do it's work, it doesn't need that connection until it's time to actually /do/ work, right? You're certainly not going to get a request processed before the servlets are loaded, so it's probably alright that the DBCP is initialized *after* the filter itself. Oh, and Yoav is right: you really want a ServletContextListener, rather than an 'init' servlet. -chris signature.asc Description: OpenPGP digital signature
Re: Symbolic links in docroot?
Hello, cd SUNWappserver/domain/domain1/docroot ln -s /usr/home/BB/public_html BB but when I try and access http://localhost:8080/BB/index.html I get 404 errors? Is this file being served by Apache or by Tomcat? (The error page should make this obvious). If it's Apache, then Apache is likely unable to serve the file for any of these reasons: 1. There's a Directory element in httpd.conf prohibiting access to files in / and others that specifically allow access to other directories. 2. The permissions on the directories leading to /usr/home/BB/public_html are not set properly (apache's effective uid or gid needs to have both read and execute permission on each of those directories). 3. Apache has been setup NOT to follow synlinks. -chris signature.asc Description: OpenPGP digital signature
Re: JNDI Datasource Reference in DD Not Necessary?
Kawthar, I've been trying to setup my mySQL connection to tomcat as well and has been reading a lot of documents on how to do this. I think I'm getting more confused. Context evtCtx = (Context) ctx.lookup(java:comp/env); DataSource ds = (DataSource) evtCtx.lookup(jdbc/mySQLDatabase); --- in my web.xml : res-ref-namejdbc/mySQLDatabase/res-ref-name --- in my server.xml: ResourceParams name=jdbc/mySQLDatabase If you have your data source defined in GlobalNamingResources then you'll need a ResourceLink element in your Context in order to use the DataSource. Other than that, things look good. -chris signature.asc Description: OpenPGP digital signature
Re: 60 second timeouts?
Allen, I am seeing pauses in page rendering that are unrelated to GC. They are pretty much exactly (60 seconds + render time) in length. Check your DNS resolver settings. I had problems like this unrelated to Java, and it turned out to be that. Make sure that localhost is actually pointing to 127.0.0.1, and that all of your nameservers are corect (and working). In my case, I had a bogus primary nameserver which had to time out before the others were tried. All networking ops sucked bigtime. -chris signature.asc Description: OpenPGP digital signature
Re: delaying context reload
Cindy, I like Tomcat's ability to reload a context when a change is made to one of my application's classes, but I'm having a problem with it reloading before the class has finished compiling. Is there a way to delay the reloading until all the files are ready? I highly recommend doing all your compilation in a 'local' directory and then using a tool like ant (I hope you're already using ant!) to copy all the updated files. This allows you to attempt a compile without disturbing a running server. (For example, if some of your files compile, but others don't you could disable your server -- even if it is only a test server). Generally, the copy will take less time that the compile ;) Hope that helps, -chris signature.asc Description: OpenPGP digital signature
Re: Help exporting a working turbine app to tomcat
Inandjo, i had a webapp developped under win2k and runing just fine. When i tried to export it to a unix machine where tomcat 4.0 is running, this is the exception i get: Horrible Exception: java.lang.Exception: Screen template '/Home.vm' not found If you're switching from win32 to UNIX, make sure that the filename is correct, including case. If you were running this on win32 already, all of your search paths should be the same, and correct. If your filenames are correct, check your TurbineResources.template file to make sure that all the / characters in your search paths are correctly done as forward slashes: / -chris signature.asc Description: OpenPGP digital signature
Re: delaying context reload
Tim, So what do you suggest in the case of copying a war file that takes 15 seconds to copy? That's a good question. I have a couple of thoughts: 1) When does the date of the inode get set -- at the beginning of the copy or at the end 2) Isn't there a way to set the time interval used to check for updates? 3) If you're hot-deploying WAR files, shouldn't you be using the Tomcat deployment tool instead of just copying it in there? ;) (I don't know anything about Tomcat's WAR deployment -- I use ant to war and unwar my apps to do hot deployment.) -chris signature.asc Description: OpenPGP digital signature
Re: Exporting turbine app to a working tomcat
Inandjo, i developped an app using TDK 2.1 (an official bundle i downloaded from apache site) on a windows 2k box, and it was working just fine. Now i would like to deploy the app on a unix machine using tomcat 4.0 with apache, by copying the folder under the webapps folder to the same folder on the unix machine. There is nothing concerning turbine on the box, just the tomcat and apache. Without commenting too much on Turbine/TDK (I've had a ... poor experience with them), I can say that removing the 'TDK' part from your application is a challenge, but doable. I was able to do it by basically starting with nothing in my WEB-INF/lib except for turbine-x.y.z.jar, and adding JAR files until it started working again. (Lots of the TDK stuff is unnecessary for 'normal' deployment). Anyhow, moving this from a working win32 machine to a Linux environment shouldn't be too much of a problem. I'd recommend divorcing the TDK (even if you keep Turbine, which is a good idea with a working app ;) but it's not cricual. You'll have to edit the configuration files that point to the TDK installation directory (there are properties like 'tdk.home.dir' and stuff like that -- just make sure they point to the right place). I would like to know what modifications i have to do in order to make the application work on that unix box (especially concerning tomcat configuration,the librairies i may have to copy to the unix box, and the changes required in the apache httpd.conf file). Well, make sure you copy all the JAR files in binary more (if you're using FTP, for example), otherwise FTP will corrupt many your binary files -- the JARs being the most obvious. You'll need pretty much everything you had on the win32 box -- including the TDK, unless you want to spend the time to break free of it and just use Turbine in a more 'standalone' fashion. So, install the TDK on the Linux box, copy your app-specific files to the Linux machine in roughly the same place they were before, and then adjust your paths in the properties files. Other than that, all you need if Tomcat and a JDK (which you mentioned were already installed). I would also like to know what is required to do for the app to work with apache. That's a whole other ball of wax. Check out the Tomcat 4.1 documentation, specifically at the Connector HOWTO. Use the JK connector. You'll have the best luck compiling it yourself against your local Apache version, so don't ask for binaries for the connector. That means you'll need a c compiler on the Linux box. That's not tough -- Linux often has gcc already installed, which works just fine with the JK stuff. Good luck, -chris signature.asc Description: OpenPGP digital signature
Re: JNDI Datasource Reference in DD Not Necessary?
All, I was just about to ask a question like this thread's today... glad I read the archives. ;) I have an intersting observation in Tomcat 4.1.29. I found that my resource-ref was misnamed in web.xml, so I decided to check it out. A long time ago, I wrote a quick-and-dirty JNDI browser to help me figure out these things. Using my JNDI browser, and looking at the path java:/comp/env/jdbc I find the following entries: * DSMstandard, class = org.apache.naming.ResourceLinkRef * diagnosis, class = org.apache.commons.dbcp.BasicDataSource (I have two DataSources configured -- I have to deal w/two DBs, so they're both there). It's odd that one of them shows up as a ResourceLinkRef and the other as a BasicDataSource. (I have no resource-ref elements in my web.xml, for testing). They were defined the same way in server.xml: GlobalNamingResources Resource name=jdbc/diagnosis auth=Container type=javax.sql.DataSource/ Resource name=jdbc/DSMstandard auth=Container type=javax.sql.DataSource/ (I'll spare you the details of the ResourceParams, but they are identical except for the name used). /GlobalNamingResources N.B.: I have my Resource declaration in GlobalNamingResources because I need to use it for my Realm. Apparently, Resources defined within a Context that will use it as a Realm don't work well -- at least this was my experience. I found that putting the resource in the global resources and then using a ResourceLink did the trick. Then, later, within the Context that will be using these, I had to add these: ResourceLink name=jdbc/diagnosis global=jdbc/diagnosis type=javax.sql.DataSource / ResourceLink name=jdbc/DSMstandard global=jdbc/DSMstandard type=javax.sql.DataSource / Why wouldn't I get both BasicDataSource objects, or both ResourceLinkRef objects? I thought it may have to do with the fact that I have the same name for the global reference and the local reference, so I changed the global references to jdbc/[dbname]-global and re-started. I got the exact same stuff in that JNDI context. Can anyone explain what's going on? Logically, I would expect the following: 1. Resource elements in the GlobalNamingResources would show up for all Contexts. If this is true, then why do I need ResourceLink in my context? 2. ResourceLink elements in a Context grant access to a globally-defined resource to that particular Context. Makes sense -- maybe you have a global resource for convenience, but it's a security issue to grant that resource to *any* webapp running on the server. So, if this is true, why don't I get two objects of type ResourceLinkRef in my context's JNDI namespace? Instead, I get one DataSource and one ResourceLinkRef. 3. resource-ref elements in web.xml should map the local name of a resource (the one you use from your webapp in JNDI lookups) to the resource mentioned in the Context's ResourceLink element. So, if all that's true, then here's what happens: Resource defines the canonical name for the resource. ResourceLink grants the resource to a particular webapp, possibly re-naming it in the process. (This should produce a ResourceLinkRef object in the JNDI namespace as seen by the webapp). resource-ref adds a 'local' reference name to a resource provided by the container. This should add another ResourceLinkRef (or something similar) to the JNDI namespace as seen by the webapp. Can someone comment on my logic, and show me where I'm making incorrect assumptions or missing some crucial detail? Thanks so much for helping me beat a dead horse. I really want to know what's going on. I'm *not* in a position where something's not working, so examples for getting things working aren't a ton of help (they pop up on the list often, and are in both Tomcat 4.x and 5.x JNDI HOWTOs). Thanks, -chris signature.asc Description: OpenPGP digital signature
Re: problem with mod_jk 1.2: error in action code
All, when you specify the connection to mysql, are you including the 'autoreconnect=true' parameter? If not, the connections will be closed by mysql after (I think) 8 hrs. I'm not sure why everyne's caught up in database connection code. This is a Coyote connector exception -- which is the HTTP/1.1 protocol connector. It allows you to connect directly to Tomcat using a browser, instead of going through Apache. Usually the Broken Pipe means the the client hung up before listening for all the data. That could mean pressing the STOP button on the browser, following another link before the page fully loads, or pressing the back button. I wouldn't worry too much about it, unless your users are somehow being affected. -chris -Original Message- From: Michael Görlich (CSC) [mailto:[EMAIL PROTECTED] Sent: Friday, March 12, 2004 9:53 AM To: Tomcat Mailinglist Subject: problem with mod_jk 1.2: error in action code Hello, I've got problems with a tomcat web application, that establishes a connection pool to a database (mysql). This connection pool is lost after an unspecified period of time the system wasn't used (f.e. 8 houres). I think the problem is due to the mod_jk connector between apache and tomcat. The web application is accessible over apache 1.3.27, that connects to tomcat 4.1.27 via jk connector mod_jk 1.2.25. The exception I got in catalina.out is: 10.03.2004 15:59:58 org.apache.jk.server.JkCoyoteHandler action SCHWERWIEGEND: Error in action code java.net.SocketException: Broken pipe at java.net.SocketOutputStream.socketWrite0(Native Method) at java.net.SocketOutputStream.socketWrite(SocketOutputStream.java:92) at java.net.SocketOutputStream.write(SocketOutputStream.java:136) at org.apache.jk.common.ChannelSocket.send(ChannelSocket.java:407) at org.apache.jk.common.ChannelSocket.invoke(ChannelSocket.java:599) at org.apache.jk.server.JkCoyoteHandler.action(JkCoyoteHandler.java:385) at org.apache.coyote.Response.action(Response.java:222) at org.apache.coyote.Response.finish(Response.java:343) at org.apache.jk.server.JkCoyoteHandler.invoke(JkCoyoteHandler.java:268) at org.apache.jk.common.HandlerRequest.invoke(HandlerRequest.java:360) at org.apache.jk.common.ChannelSocket.invoke(ChannelSocket.java:604) at org.apache.jk.common.ChannelSocket.processConnection(ChannelSo cket.java:562) at org.apache.jk.common.SocketConnection.runIt(ChannelSocket.java:679) at org.apache.tomcat.util.threads.ThreadPool$ControlRunnable.run( ThreadPool.jav a:619) at java.lang.Thread.run(Thread.java:534) 10.03.2004 15:59:58 org.apache.jk.common.ChannelSocket processConnection INFO: server has been restarted or reset this connection The HTTP connector defined in tomcats server.xml is: Connector className=org.apache.coyote.tomcat4.CoyoteConnector port=8009 minProcessors=5 maxProcessors=75 enableLookups=false redirectPort=8443 acceptCount=10 debug=5 connectionTimeout=0 useURIValidationHack=false protocolHandlerClassName=org.apache.jk.server.JkCoyoteHandler scheme=http secure=false / The log in mod_jk.log shows the following: [Wed Mar 10 15:59:58 2004] [jk_ajp_common.c (1052)]: ERROR sending data to client. Connection aborted or network problems [Wed Mar 10 15:59:58 2004] [jk_ajp_common.c (1303)]: ERROR: Client connection aborted or network problems Here is the summary of my environment options: - tomcat 4.1.27 - apache 1.3.27 - mod_jk 1.2.25 - Suse Linux 7.2 - Java 2 RE Standard Edition, Version 1.4.1 (build Blackdown-1.4.1-01) My special problem is, that I'm not able to reproduce the error, it occurs after an unspecified period of time. I found nothing related to this problem. Any ideas? Thanks in advance M.Goerlich - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] signature.asc Description: OpenPGP digital signature
Re: reclaiming memory problem PS
Jerald (or is it Gerald -- your email address and 'name' don't match), Now sessionStatus is getting caught fine, but when I try and redirect to a JSP after that, nothing happens. I originally tried mapping.findForward (Struts), response.sendRedirect and forwarding using RequestDispatcher. I have tried getSession(true) and false. What implications (if any) does session timeout have in terms of forwarding after the session is invalidated? Session state should have nothing to do with your ability to forward, etc. Can you post thesnippet of code where you try to redirect the user? -chris signature.asc Description: OpenPGP digital signature
Re: Serialization issue
Sean,
I'm trying to grab the current session ID and the request parameter map
from the current request and serialize them into a base64 string to
pass to a PHP application. Problem is, I keep running into the
following exception everytime I try to serialize anything imlementing
the Map interface:
java.io.NotSerializableException: org.apache.coyote.tomcat5.CoyoteWriter
If you're passing this to PHP, I'm guessing that you don't want to use
serialization. (Unless PHP can de-serialize Java objects and I didn't
know...)
I think you want to do your own type of 'serialization', something like
this:
Map parameters = ...;
StringBuffer sb = new StringBuffer();
for(Iterator i=parameters.entrySet().iterator(); i.hasNext(); )
{
Map.Entry entry = (Map.Entry)i.next();
ab.append(entry.getKey())
.append('=')
.append(entry.getValue());
if(i.hasNext())
sb.append(',');
}
// send sb (or sb.toString()) to PHP
Hope that helps,
-chris
signature.asc
Description: OpenPGP digital signature
Re: RegExp issues with Tomcat 5
Karl,
Using the regexp tag library I think it is still at version 1.0.
Because INSERT into tablename (field1) values ('I CAN'T DO THIS')
generates an error in SQL Server.
If you use a PreparedStatement, then you can send any string to the
statement object and it will do it's own escaping.
-chris
signature.asc
Description: PGP signature
signature.asc
Description: OpenPGP digital signature
Re: Global URL Redirect Tomcat 5
Nathan, I am trying to migrate to Tomcat 5 from a different app server. Is there a way to define url redirects? i.e. if a page moved off your server to a different location and you do not want to make a redirect page. You could write a servlet that's mapped to /*, and then have that servlet replace the hostname (and part of the path?), then send a redirect back to the browser. Would this meet your needs? -chris signature.asc Description: PGP signature signature.asc Description: OpenPGP digital signature
Re: ClassNotDefError problems within JAR files under Tomcat 4.1.12
Nathan, I am developing an imaging servlet under Tomcat 4.1.12 using JAI 1.1.2. Every time I update my code to add new features, it will return with a NoClassDefFoundError until I restart Tomcat. At that point, it finds the 'missing' class and everything works as expected. I'm no expert on JAI, but I believe that it uses JNI and a native library to do some of it's dirty work. If that's the case, then you should make sure that the JAI JAR file and native library are loaded by a classloader outside (higher than) your webapp. There's documentation on the Tomcat site (and probably other servlet containers, too) that says that native libraries should only be loaded one time. IF they get loaded multiple times (as would happen if they were loaded by the webapp), strange behavior can result. Try putting jai.jar (or whatever) into TOMCAT_HOME/common/lib and jai.so (or whatever) in a convenient place where it can be found. (Sorry, dunno where that might be. Anyone else?) Hope that helps, -chris signature.asc Description: PGP signature signature.asc Description: OpenPGP digital signature
Re: reclaiming memory problem PS
Jerald, session.setMaxInactiveTimeout(-1); Yeah, this is a bad idea. The session will never go away by itself. This *requires* the user to press a logout button, and for you to explicitly call session.invalidate(). Users frequently do not log themselves out, and their sessions will never die. You will eventually run out of memory. If you need a long timeout, just make it really long (like a couple of hours). There's usually no good reason to make it -1. PS is the session time out linked wirth inactivity? My session attribute only persists as long as I am using the app. That's exactly how the 'inactive' timeout works. -chris signature.asc Description: OpenPGP digital signature
Re: OracleConnectionPoolDataSource creates too many connections
Rudi,
I have two things to add that nobody seems to have mentioned.
In my LoginServlet, I create a new PooledConnection, which I add to the
Servlet Context:
PooledConnection pc = ConnectionFactory.getInstance.getPooledConnection();
ServletContext ctx = getServletContext();
ctx.setAttribute(pooled_conn, pc);
Gah! Why are you putting a database connection into the application
context? This sounds like a concurrency nightmare! I think you want to
put the /pool/ into the application scope, not a single connection.
Connection con = pc.getConnection
//do a few things with the connection
try{
if (con != null){
con.close();
} // I'm assumingthis returns the connection to the Pool
}catch(SQLException sqle){
sqle.printStackTrace();
}
You need more try/catch blocks. conn.close should /always/ be in a
finally block:
Connection conn = null;
// Declare your statements and resultsets, here, too
try
{
conn = // whatever
// Do stuff with connection
}
finally
{
// Close your statements and result sets, here, too
if(null != conn)
try { conn.close(); } catch (SQLException sqle)
{ /* log this exception somewhere */ }
}
Hope that helps,
-chris
signature.asc
Description: OpenPGP digital signature
Re: how to get or build a mod_jk module for Cygwin Apache ?
Flo, i try to configure Apache 1.3.xx from cygwin and Tomcat 5.0 from windows to work together with JK. Okay. Isn't there a win32 binary? http://apache.towardex.com/jakarta/tomcat-connectors/jk/binaries/win32/ (Look for files with 1.3.27 in their name -- those are for Apache 1.3.27). But i have a problem getting a binary version of mod_jk. At this URL : http://apache.crihan.fr/dist/jakarta/tomcat-connectors/jk/binaries/ the freebsd directory is empty. Maybe I'm confused. Why are you looking in the freebsd directory for cygwin binaries? For cygwin i don't know if i must use a win32 mod_jk.dll (version 1 or 2) or build a mod_jk.so. It will be a .dll file. However, I'm not sure cygwin has anything to do with this. You'll be running Apache-win32 and mod_jk-win32, and I assume Java/Tomcat will be running on JDK-win32, so where does cygwin come in? -chris signature.asc Description: OpenPGP digital signature
Re: RE : how to get or build a mod_jk module for Cygwin Apache ?
Flo, I thought FreeBSD works like cygwin but it seems wrong. FreeBSD is a 'real' UNIX flavor, while cygwin provider UNIX-like services and libraries on win32. I tried the mod_jk_1_2_5_2_0_47.dll I got errors launching httpd : Win32 error 126 (The specified module could not be found) I needed to add ApacheCore.dll Win9xConHook.dll to the PATH Then i got another error message : Win32 error 127 (The specified procedure could not be found). You shouldn't have used mod_sk_1_2_5__2_0_47.dll -- that's for Apache 2.0, not Apache 1.3. Then I tried mod_jk2-1.3.27.dll but I get this error : Syntax error on line 1020 of /etc/apache/httpd.conf: Can't locate API module structure `jk_module' in file /usr/lib/apache/mod_jk2-1. 3.27.dll: dlsym: Win32 error 127 You probably have the dll in the wrong place. The standard location for Apache modules is in the 'modules' directory in the Apache installation. So I thought to build any mod_jk but how? You'll have the same problems if you build it yourself. You still need to put it in the right place. I would like to find a better and easier solution. Are you using Apache as a package that was installed via Cygwin? You might have better luck with the 'standard' distribution, which comes with a very simple installer, from httpd.apache.org. -chris signature.asc Description: OpenPGP digital signature
Re: Virtual Hosts with Apache and Tomcat
Antonio, In fact, cap it at 10 and watch the app dring to a halt before it even gets going. This is a pretty compelling example. If the pool is drying up, they're definately screwing up. It is. But developers may reply: You are using less connections than those specified in (the contract) / (the manual) / (fill in yourself). I thinks we're misunderstanding each other. I think that when the pool is capped, and the connections are never returned, you get to a point where the pool refuses to give you a new connection, no matter how long you wait. This is a pretty good idea for some basic debugging. You should only have to demonstrate to your devs that you can deadlock their server by capping the connection pool. After that, it's their problem, right? :) With the proposal, you demonstrate they have a connection leak, which is the real problem. Once you showed them they had ONE connection leak, you can urge them to dig for other connection leaks themselves. But, of course, the idea about the deadlock seems really good also. If I understood, what you mean is: If you set the connection pool size too low for the app, it should crash at will (or better, show an 'unavailable' screen), but it should continue working as soon as load permits it. Am I wrong? I'm thinking that the connections are added to the pool upon request (from his observationa, it looks like the 10 pre-allocated connections are always ignored), and then never returned. The pool remembers the pre-allocated ones, plus the ones it created on-the-fly. I think that if he caps the connection pool size at 25, it will only take 20 requests to lock up the server for want of a DB connection. Try setting the pool size to 11 and see if you lock up after one request ;) -chris signature.asc Description: OpenPGP digital signature
Re: Process Died | Production
Yoav, I've had the displeasure of experiencing a Tomcat JVM core dump on me not too long ago (in a multi-user development environment, fortunately) on a Sparc/Solaris box. It turned out to be due to insufficient swap memory in the system. I didn't see any OutOfMemoryExceptions raised prior to the crash, so it wasn't that obvious, but the top command came to the rescue in the end. Good one ;) It reminds of another time, where I had internal crashes occurring due to a low limit of file descriptors (ulimit -a will show you the limits). That one was frustrating ;) I'm sure many people on this list of horror stories with seemingly random crashes... I've experienced even more random crashes (SEGV). It turned out to be bad memory (or bus), and it only showed up under pretty heavy load. :( -chris signature.asc Description: OpenPGP digital signature
Re: Securing SSL from IIS to Tomcat
Antonio, Nice demonstration. I only miss the seriously part, which is covered partially by the term expensive ;-) Right. :) I wanted only to add IPSEC as another possible scenario. It has less (no) TCP overhead than a SSL tunnel. And it works without even changing your config files. Now that's an idea. I don't even know how IPSEC works. However, I think it might be a UNIX-only thang (and he's on win32). However, the point you make about VPNs is a good one. I think you can't 'avoid' the TCP overhead, though, since this stuff runs over something which has to have some kind of error-correcting and delivery guarentees like TCP does (even if it's not actually implemented using TCP). At any rate, you could tunnel your JDBC connections through some other kind of secure connection. -chris signature.asc Description: OpenPGP digital signature
Re: Virtual Hosts with Apache and Tomcat
Stephen, In fact, cap it at 10 and watch the app dring to a halt before it even gets going. This is a pretty compelling example. If the pool is drying up, they're definately screwing up. Whoa there pardner: I am not going to deliberately cripple a production box. The problem has been demonstated in test environments and that is as far as I will intentionally let it go. Oh, I totally meant in a development setting. I would never suggest that you cripple a production box. You can easily demonstrate the problem. Do that in dev, and make them fix it. Then, deploy the fix as part of your regular deployment procedure. -chris signature.asc Description: OpenPGP digital signature
Re: Virtual Hosts with Apache and Tomcat
Kennan, I can agree partially to yours. But if you see him, he doesn't know about the impact of JVM and tuning parameters, as he mentioned in his email. Do you expect him to take a lead in fixing that? I have seen the projects losing its focus by the nature of peoples deviating to get their interests fulfilled. This is not a 'special' interest. This is a legitimate resource leak that he wants them to fix. He can demonstrate the leak. That's all that's necessary on his part. The rest is up to the developers. I'm not suggesting that he fix the problem. Only to demonstrate it and get the developers to fix the problem. I would appreciate, if the developer and sysadmin working together in this problem (i doubt verymuch as sysadmin involvment, all he can do is give top or sar reports). Sysadmin has much knowledge in configuring servers, architect the infrastructure, manage the network, backups etc. Yes, but *this* sysadmin also has enpirical data that demonstrates the resource leak. Forget sar and top. How about the app locks up. That should be motivating enough. I never seen any sysadmin trying to fine tune any Application Servers. Actually, the sysadmin is the /perfect/ person to fine-tune app servers. Most devs don't know jack about the app server they use. That's why they deploy onto app servers with standard interfaces and services (Servlet and JSP spec). The deployment and admin folks are the ones who should know how to configure the app servers. If that is the case, then the project sucess will be in stake. Everyone has to do their own roles. If I would be the sysadmin, then i would tell the developers to go these newsgroups. Dont you think that most of developers resolve their issues by newsgroups and websites for their problems. Here's the problem: the devs refuse to admit there's a problem. They won't go to the newsgroups to ask about a problem that they don't believe exists. That's why the sysadmin is here. He wanted to get some information on how to prove that there's a leak. He's gotton that information. Let's wait for the devs to visit the group, now ;) He clearly mentioned that the developeers raised that questions and trying to get the verification from the newsgroups. Dont you think that is the part of communication gap between the developers and him. If he is very keen, why not one of the developers responding his thread and get the issues fixed for the project. I think the problem is that the devs think the sysadmin is foolish and wrong about the resource leak. Now that he can demonstrate the leak, they will take him more seriously. I believe that we have helped in this situation, and that the devs will now address the problem instead of sticking their heads in the sand. -chris signature.asc Description: OpenPGP digital signature
Re: Deploying a simple servlet
Leon, I'm new to tomcat and am trying to run a very simple hello world type of servlet. I've configured the web.xml to mimic a similar servlet in tomcat examples. Still while all the example servlets from tomcat work mine fails to run. I receive the following message: HTTP Status 404 - /MyApp/servlet/Testing I would greatly appreciate any advice. In the newer versions of Tomcat, the 'invoker' servlet is disabled by default. The best way to get your servlet running it to create a servlet-mapping element in your web.xml and then use your mapped URL to access the servlet (instead of using a URL like /context/servlet/ClassName). -chris signature.asc Description: OpenPGP digital signature
Re: RegExp issues with Tomcat 5
Karl, We upgraded tomcat running one of our applications from 4 to 5 this past = Friday, March 27th. Since, one of our regular expressions is not = functioning properly. Here is the expression: s/(\\|')/\\$1/g Instead of escaping the apostrophe with another apostrophe for SQL = purposes, it is replacing the apostrophe with the literal '$1'. It was = working with Tomcat 4 and that is the only thing that has changed. What regex library are you using? Tomcat shouldn't have anything in it that would affect your regexs. Not to open up too big of a can of worms, but if you're using JDBC, why do you have to escape your own SQL statements? -chris signature.asc Description: OpenPGP digital signature
Re: Deploying a simple servlet
Leon, I've set up a directory structure for this app exactly like the servlet examples structure and after making changes to web.xml I'm recieving the following root cause error: java.lang.NoClassDefFoundError: TestingServlet (wrong name: myApp/TestingServlet) Is your TestingServlet class in a package? If so, you need to give the fully-qualified class name in your web.xml file. If it's not in a package, I've heard many reports that it simply won't work :( Try putting your servlet class into a package, updating your web.xml file and trying again. -chris signature.asc Description: OpenPGP digital signature
Re: scope order of initialization of variabes includes
Marc, I thought I should get the output: First page: ok Second page: ok I think this is what's happening. (You can probably verify this by looking at the java source code generated by Tomcat from your JSP): When you do a jsp:useBean and declare where the bean lives, the translator essentially converts that to this line of code: String bandBg = (String)request.getAttribute(bandBg); Then, later, you are using a scriptlet to do this: bandBg = ok; In the other JSP, the same code executes: String bandBg = (String)request.getAttribute(bandBg); However, you never poked 'ok' back into the request, so it's still either null or blank or whatever it was when you started. Setting the local reference bandBg in your Page1.jsp doesn't do anything outside of it. You'd have to do something like this in your scriptlet to make it work: % request.setAttribute(bandBg, ok); % If you're going to use jsp:useBean, you probably shouldn't screw around with those variables inside of scriptlets. Try to stick with the bean tags only. That generally means that you'll have to write your own bean instead of using a string (so it can have a 'setValue' method or something). Hope that helps. -chris signature.asc Description: OpenPGP digital signature
Re: [OT] Re: Process Died | Production
Ankur, I've experienced even more random crashes (SEGV). It turned out to be bad memory (or bus), and it only showed up under pretty heavy load. :( This is a little OT, but just out of curiosity, has anybody been successful in gaining root/tomcat/whatever-uid shell by capitalizing on a JVM's (not necessarily tomcat's) core dump? I've always wondered if that was possible. I know its extremely hard (impossible?) to consistently overflow JVM's stack, but has it ever been done? I've never heard of anything like this before. However, Java's stack is not what gets overflowed, here. IF the JVM goes down, it's the JVM code that faults, not the Java code itself. Java's stack and heap are pretty far away from anything that's executing. Generally stack or buffer overflow bugs are exploited because someone corrupted the stack and not only inserted some maliscious code, but also modified the return pointer to call that code. In Java (without native code, at least), I don't think you could intentionally do anything like this. I suppose it you had an evil app connect to a JVM using something like JVMPI, you might be able to do something like this. However, you usually don't have the profiling hooks turned on in production ;) -chris signature.asc Description: OpenPGP digital signature
Re: reclaiming memory problem
G, But, when I close down a client window, the memory consumption remains at it's peak (and climbs with each successive new client opened) In other words, memory consumption increments but never decrements, necessitating a server bounce at unacceptable frequency. What is your definition of 'never'? Java cannot immediately reclaim the memory used by that user's sesson if the user closes the browser. Tomcat doesn't know that the client has closed their browser. It must wait until the session times out (30 minutes, I think) before the session can be cleaned up automatically. Solutions? 1. Reduce the session timeout. 2. Store less stuff in the session so that it's not such a memory drain. 3. Beg your customers to logout before leaving your app. ;) Why are the resources not being reclaimed, and how might I gently encourage resource reclamation? I fear System.gc() is not the answer here. System.gc only tells the GC 'now might be a convenient time to run the GC'. Other than that, it's not much good. -chris signature.asc Description: OpenPGP digital signature
Re: Virtual Hosts with Apache and Tomcat
Antonio,
And bad. Every time I restart, Tomcat loses the state information for
established login sessions. Customer don't like that.
That (with a high probability) is because some objects they store in
sessions are not Serializable. IOW, they violate the Servlet Specification.
I'm just curious: is this actually a violation of the servlet spec? The
API seems to indicate that you can put anything in the session that you
want. I don't think it has to be Serializable... thought I was wrong
before, once ;)
Tried that. Capped it at 35 and the webserver stopped servicing any
DB request as soon as the pool reached 35. This is why I believe the
pool management is faulty and/or something is hogging all the
connections.
I share your belief. Let's try to prove it. Raise it to some other
figure, and see if the same happens again. Ask them how big should the
figure be.
In fact, cap it at 10 and watch the app dring to a halt before it even
gets going. This is a pretty compelling example. If the pool is drying
up, they're definately screwing up.
Oracle 9i takes 16M per connections. So Oracle claims. I've tested
it as high as 20M. I generally use 18M as a guideline
I've heard (not a DBA, though) that Oracle 9i has a mode where it does
not spawn a process per connection, but uses threads instead (?) and in
that mode it uses far less resources. This way, we have some modest
Oracle servers hjandling up to 300 simultaneous (mostly idle) connections.
It depends on the size of your rollback segments and the number of
transactions you are doing. If you do big transactions, each DB
connection (thread *or* process) wioll need a big chunk of memory. I
wouldn't kill yourself trying to figure out how to reduce this process
size. Fix the real problem, which is poor connection management.
I'll mention DBCP and see what happens
DBCP has a nice removeAbandoned feature.
Otherwise, you can use this code (tweak it to your needs) to track where
connections are opened and closed from:
(code not tested at all)
// open method signature
// code that opens the connection (and stores it in conn variable)
try {
throw new Exception(Pool Debugger says: Connection + conn +
opened:);
} catch (Exception e) {
e.printStackTrace();
}
I've seen code like this before. Many people think you can't get a stack
trace unless you throw an exception. Not so. All you have to do is
instantiate it, and you get the stack trace. So, the following will
produce identical results, without the nasty try/throw/catch:
new Exception(Pool debugged says: ...).printStackTrace();
I would recommend explicitly printing out the hashCode of the Connection
object itself, just in case the connection doesn't include any
identifying information in it's .toString method.
Then you can...
#!/bin/sh
# Filter pool debugger statements.
This is a pretty good idea for some basic debugging. You should only
have to demonstrate to your devs that you can deadlock their server by
capping the connection pool. After that, it's their problem, right? :)
-chris
signature.asc
Description: OpenPGP digital signature
Re: Virtual Hosts with Apache and Tomcat
Kannan, Being yourself as SYSADMIN for UNIX and Network, it would be nice that developers or professional should take a lead into get into this problem. Easy for you to say. Let's face it: these guys have a connection leak. Plain and simple. Your devs need to find their leak. It is demonstrable. It locks up the server. QED. Make them fix it. This isn't about communication or a sysadmin whining to the devs about something he doesn't understand. This is a resource leak. It is apparently well-understood. He's done his homework. They are clowns. -chris signature.asc Description: OpenPGP digital signature
Re: Syntax Question
Jack, I want to add a variable to the web.xml file and be able to call in from within a servlet, taglib or bean. You need to use the init-parameter element in your web-app element in web.xml. Your JSP syntax appears to be correct. -chris signature.asc Description: OpenPGP digital signature
Re: Do servlet threads ever die?
Ken, Problem: I call native code (Tcl) from my servlet. Tcl's thread model forces me to run Tcl only on the thread that created the Tcl interpreter. So now how do I cleanup these interpreters? The cleanup code needs to be run on the creating thread. Guh. I highly recommend against running native code from an app server. There are lots of reasons I could enumerate if you ask. Number one reason: crashed native code = crashed server. :( Is there a way you can call the Tcl code through a socket or by invoking another process and communicating with it via stdin/stdout? (see java.lang.Process) Do servlet threads ever die besides during shutdown? If not, I may have no problem. AFAIK, Tomcat never retires threads unless something horrible happens (or tomcat is shutting down). Can I register a listener to be invoked on the thread before it shuts down? I don't think you can do this to a thread. You might be able to hack Tomcat's thread pool so that you can cleanup after a thread if it's going to be retired. I don't have any good ideas. Sorry I couldn't help more. -chris signature.asc Description: OpenPGP digital signature
Re: Securing SSL from IIS to Tomcat
Antonio, Since IIS decrypts the request and passes it unencrypted to Tomcat, How do I encrypt the request so that all communication from IIS with Tomcat is secure??? Do you really need it? It would probably affect your performance seriously. (No, I cannot prove that statement...) I can prove this statement. :) Connection latency = network transfer time + 2 * (TCP overhead + encryption/decryption) (Twice since both clients have these overheads). When the encryption/decryption term becomes zero (removal of SSL), performance increases. The assertion that encryption/decryption takes more than zero time is left as an exercise to the reader. (Oh, and SSL acceleration hardware is expensive, which would seem silly if the ops were trivial). Latest reference I've been able to google for is from Tomcat 4.1 JK connector (deprecated). But I have not searched a lot. I don't think that the JK mod is deprecated. In fact, it's suggested for production installations with Apache 1.3 and 2.0 (and IIS, I guess). (From the other followup post): I doubt any has that kind of need. If that is the case, then everyone will pull their brains out. Often, there's a need. What if you don't trust your own network? If we are worried about the IIS having SSL with Tomcat, then how about SSL between database and Tomcat. How about query tools used in database ( secured query tool? , I dont know who has like this). Some people need this type of stuff. Let's say that your database server is on another network, and you need confidential transfers? I would not worry, if the Tomcat behind firwall, and behind the IIS (SSL). I would still worry. For example, all of our hardware is hosted by a data center. We get all kinds of M$ virus crap spewing into our syslogs every single day. We get SMB traffic logs. We get all kinds of crap. There's nothing stopping me from running a packet sniffer on my machines, which would compromise the entire network if SSL were not being used. (Fortunately, each app is small enough that we can runn both the app and db on the same machine and avoid those security concerns). The JK documentation says nothing about SSL (other than forwarding SSL information from thr web server to Tomcat). I assert that you have two options: 1. Use Tomcat standalone (why not? do you have a huge amount of non-dynamic content? If not, consider Tomcat standalone) 2. Set up an SSL tunnel using your favorite ssh software. OpenSSH is available for win32 (at least through cygwin). You can punch through pretty much any network and firewall (as long as they allow SSH). All the data is encrypted. Most ssh clients/servers also support compression as well, which /might/ help offset the performance penalty of encryption (due to lower network transfer times). Good luck. -chris signature.asc Description: OpenPGP digital signature
Re: The server encountered an internal error () that prevented it from fulfilling this request.
Metin, What can be the reason of this error? java.lang.NoClassDefFoundError: org/jdom/JDOMException Looks like you need to install the JDOM library... -chris signature.asc Description: OpenPGP digital signature
Re: Tomcat 4.1.29 JVM crash on RHEL-3-ES
Joe, I'm experiencing sporadic crashing of the JVM running Tomcat in 2 of my environments. These environments both run the same webapps, and they both use JDK 1.4.2_03 on RHEL-3 ES. Other environments where these same webapps run are using JDK 1.4.1 and Redhat7.2, and are not experiencing these crashes. What kernel version are you using (I assume you're using 2.4.something). There's a lot of talk on this list about having to set the LD_ASSUME_KERNEL environment variable in order to prevent crashes like these on various versions of Linux (esp. RedHat). I myself have never needed this setting. Do a quick search in the archives for LD_ASSUME and see whet they have to say. -chris signature.asc Description: OpenPGP digital signature
Re: classloader order
Yoav, What can I do to make sure that whatever I have in my /WEB-INF/lib/ directory takes precedence? You're not doing anything wrong, your systems group is. You can't override the bootstrap classloader. Aw, sure you can (-Xbootclasspath/p). But trumping the boot classloader with WEB-INF/lib is an even worse transgression than putting app-specific JARs into JAVA_HOME/lib/ext. -chris signature.asc Description: OpenPGP digital signature
Re: Virtual Hosts with Apache and Tomcat
Stephen, I am having a problem with tomcat opening up up a number of connections to an oracle server that never get closed. This causes the number of open connections to build up over time and, eventually, causes the oracle server to use all of its swap. That's not good :( Restarting tomcat clers this up. That's good! :) I think there is a problem with some jsp's opening connections and then not closig them but the developers claim (surprise) their code is clean. It's tough to make sure that database connections (and statements, and result sets) get cleaned up in JSPs, unless you have a talented JSP author. (Most JSP authors aren't that talented, unless they are also good Java developers, in which case they would have implemented the DB access in a servlet and just used the JSP for display. Anywho...) If the number of connections keeps going up and never tapers off or stops altogether, then something is misconfigured with your connection pools. Even if the engineers say that the pages are clean, you should protect the app server (and the DB server) from being swamped by capping the number of DB connections allowed. Ever. Any decent DB connection pool lets you specify this kind of thing. You should set that value to something reasonable. You can get away with a suprisingly low number of these. (I was consulting on a big project that was a somewhat DB intensive, web-based app. They had the app server configured to accept 75 simultaneous connections. They also set the db connection pool size to 75. I asked why and they basically said so that every HTTP connection can get a db connection. Duh. I talked to management and make them put in debugging information to find out how many connections were ever in use simultaneously. Seven. (Suckers). They also didn't realize that Oracle takes like 10MB per connection on the backend, and they had six physical app servers running two separate copies of the application. That's 75 * 6 * 2 * 10MB = 900MB. Good thing the DB server had 3.5GB of RAM, but still...) The explanation they give is: The increase in number of connections beyond the CACHE_MAX_SIZE setting in the app1.properties file is due to the private labeled sites. For each virtual host (private labeled site), there will be a separate JVM running the Tomcat web server space. For each of these JVMs, there will be a separate database connection cache pool to serve the user requests. This is the designed functionality of a web server that will support virtual hosts. I don't know tomcat near as well as I do Apache but this sounds like someone is blowing smoke. This isn't too outrageous, actually. If each webapp has its own connection pool, and they are configured to have at maximum, say, 10 connections, then you'll probably end up with 10 * webapp_count connections to the database server, regardless of the number of Tomcats/JVMs are running. If Tomcat is configured to handle the connection to the database (say, through a Realm and a JNDI-configured connection pool), you might be able to share connections between all of the webapps. If you solve the private-labelling problem by using multiple webapps, but through the same database, this approach seems like an excellent idea; configure Tomcat to provide a JNDI-based connection pool, and then configure the separate applications to use that pool. That way, you can control the total number of connections across all private labels, instead of having them be independent. If I run ps on the server it looks to me like there is only one instance and if I restart tomcat, _all_ virtual hosts are restarted. Yeah, then it's definately separate webapps running on a single instance of Tomcat. Try to pitch the above idea to your engineers and see what they say (probably something like it's fine the way it is!). I don't care who is right or wrong but I do want to clear up this problem. Any ideas? If you need any more information, just ask. I think I'd need to know if the connections were really never going away. Use netstat to find out what state they're in. If they all say ESTABLISHED, then you've got a connection leak. If many of them say TIME_WAIT or something like that, then you might have a problem with either the client or the server not properly hanging up the phone. If it's the former, then yell at your engineers. Cap that connection pool size at something reasonable, like ten connections. After that, the application starves. That's good for the app server and the database, while bad for your application. You can use Jakarta Commons' DBCP as your connections pool. It has some wonderful debug options, like giving you a stack trace for the code that obtained the connection if that connection isn't returned within a certain amount of time. That can save days or weeks of code reviews. If your connections are in TIME_WAIT, see how long they stay that way. Waiting 5-10 minutes for a connection like that to get
Re: image not displayed
Peter,
Since i am new to Tomcat i installed tomcat 5 on a Suse 8.2 linux machine and tried to install the First Webapp Servlet example.
I have placed all files in subdirectories of myapp, thus:
myapp/src/mypackage/Hello.java,
myapp/web/WEB-INF/web.xml
myapp/web/image/tomcat.gif
After ant install and starting a local browser, the text output is shown correctly but the image is not displayed. I cant figure out why the image isn't shown. Did i mis something in the web.xml? (see below)
web.xml is probaly not your problem. What does your 'ant install' do?
Are you using Tomcat standalone? If so, your context path (in
server.xml) should point to the directory where your WEB-INF and 'image'
directories are. If you are using Tomcat's auto-deploy feature and just
shove your files into {tomcat-home}/webapps/myapp, then make sure your
image path is {tomcat-home}/webapps/myapp/image. Also, make sure that
your URLs for the images in your web page look like this:
img src=http://host/myapp/image/tomcat.gif; /
or
img src=/myapp/image/tomcat.gif /
...or they are relative to your actual web page or JSP. I recommend
against this, since you then get lots of img
src=../../image/whatever.gif / which gets unmanageable after a while,
especially if you start including some JSPs in others where the paths
aren't always the same relative to the root.
-chris
signature.asc
Description: OpenPGP digital signature
Re: Help tomcat problem with memory
So, Hi this the statemens that i posted in the catalina.sh file JAVA_HOME=/usr/java/j2sdk1.4.0 ; export JAVA_HOME CATALINA_HOME=/usr/local/tomcat1 ; export CATALINA_HOME JAVA_OPST=-server -Xms30m -Xmx40m -Dfile.encoding=ISO-8859-1 CATALINA_OPTS=-Xms30m -Xmx40m Your app seems to be taking more than 30m, eh? Try 30M -- it might be case sensitive. Although the VM and Tomcat might not even be able to load and get anything done in that little memory. This the output of the top command and both tomcats has been installed in the same server 1:22pm up 21 days, 5:36, 6 users, load average: 0.17, 0.23, 0.28 Your load average looks great. What's the problem? CPU0 states: 9.5% user, 4.0% system, 0.0% nice, 85.5% idle CPU1 states: 7.4% user, 8.5% system, 0.4% nice, 83.3% idle Mem: 1030580K av, 1019220K used, 11360K free, 0K shrd, 92348K You've got a dual-CPU machine with a gig of RAM and your app is responding slowly? What are you running? buff Swap: 2096376K av, 58748K used, 2037628K free 706776K Your swap hasn't been touched. Unless you have a totally braindead app, something is terribly wrong if it's really not responding. Have you hooked up a profiler to your app? -chris signature.asc Description: OpenPGP digital signature
Re: Help tomcat problem with memory
Fabian, If i have only two java proccess of 141 MB and 55 MB why my memory is over 90% of the utilization and the application working slowly or not working, when the users try to get access using the webserver in this server. i can't understand this problem...thnaks You have to understand something about Linux memory: it never gives it up. So, even though 'top' reports that your box has very little free physical memory, that doesn't mean that it's all being used. When the kernel obtains memory on behalf of a process, it doesn't go back into 'free' when the process fees it or dies. I know it's a little weird. However, you do have a reasonable complaint: the app seems sluggish, and with that hardware, it shouldn't be. Again: what are you running? If your code fires off 500 threads every time a request comes in, then maybe you shoudl rethink your architecture. If it's nothing heavy-hitting, then something else is wrong. What else is running on the box? What benchmarks have you run? Or, are you just complaining about wall-clock time? Could your problem be network latency? How far from the box are you? -chris signature.asc Description: OpenPGP digital signature
Re: Tomcat heap vs. java.exe
Asim, kill doesn't work on Windows. Besides, I can't afford to stop the server right now. Many of our customers are using currently using the websites. In order to get a thread-dump, you can do a CRTL-BREAK (not CTRL-C) on win32. Unfortunately, you have to already have started the JVM on a console, instead of as a service. If it's already running, and you can't restart, you can write a JSP that will get a thread dump for you. Just look up Thread.getThreadGroup, ThreadGroup.getParent, and ThreadGroup.list and you can throw something together pretty quickly. Ideally, you can get a VM thread-dump, though. Any idea why JVM would run out of memory? I thought GC would clean up the memory at a certain ratio. In general, yes. However, it's easy to eat-up memory when you think the GC will take care of *everything*. Sometimes, you will get a legitimate OutOfMemoryError. These occur when your application just plain needs more memory than you have allowed Java to use. Setting the max (and usually min) heap size to something larger will certainly help with that. However, if you have a memory leak, it will only prolong the enevitable. -chris signature.asc Description: OpenPGP digital signature
Re: Tomcat heap vs. java.exe
Hello, If I were you, I'd load test your site first before going to production. As you pointed out, you cannot afford to stop your production server or hang it with OutOfMemory errors. This is good advice. Load testing is generally easy to do in development, and hard to recover from if you don't do it, once you're in prod. You can get jvm dumps on windows by turning on the -verbosegc. This will dump out stack traces repeatedly without exiting the JVM. I don';t think this is accurate. '-verbosegc' only logs the GC activity to the stdout or stderr or whatever. It does not generate stack traces for you. CTRL-BREAK on win32 gives you a full thread-dump. This will give you a stack trace for any running thread, as well as information about it's state (runnable, waiting on monitor, sleeping, etc.). This can be valuable when your application is not responding in timely way. If you're thrashing or getting LOTS and lots of GC activity, your memory settings might be off. On the other hand, if you have horrible code, you can never recover from that, even with a dual-cpu box ith a gig of ram :) -chris signature.asc Description: OpenPGP digital signature
Re: Run Applet on Tomcat 5.0
Henry, Is anyone know where is Web-accessible directory in Tomcat 5.0.16? A web-accesible directory is anywhere you could put a JSP or HTML file or image. Just some place where a URL can point to the class file and the server will give the browser access to it. -chris signature.asc Description: OpenPGP digital signature
Re: Run Applet on Tomcat 5.0 -- Where to put the plugin class files?
Henry, Please respond to the list, so everyone can see this conversation. Also note that this isn't really a Tomcat question. It's an applet question. However, I'm happy to help. 1. I put PluginApplet.jsp in the directory [TomcarInstalledHome]\webapps\ROOT\ Okay, that should be fine. Since IE is trying to load your applet, I assume that the page is in the proper place. 2. I use IE to browse http://localhost/PluginApplet.jsp Result: IE failed in initializing PluginApplet. Where should I put the following 4 class files? DrawingPanel.class ,PluginApplet.class, TextPanel.class, WindowUtilities.class First of all, I highly recommend using a JAR file since you have multiple class files. I believe that most browsers will use the current base URI for loading classes, so your classes should be in the same directory as your PluginApplet.jsp file (or, at least, the URL created by replacing 'PluginApplet.jsp' with 'PluginApplet.class' should resolve to your class file. jsp:plugin type=applet code=PluginApplet.class width=370 height=420 /jsp:plugin You should, however, specify either the 'codebase' attribute for the jsp:plugin element, or use the 'archive' attribute, which would be required if you were to use a JAR file. So, if you JAR'd all your class files together into, say, PluginApplet.jar, then your JSP code would look like this: jsp:plugin type=applet code=PluginApplet.class archive=/path/to/PluginApplet.jar ... other attributes like width, etc. ... /jsp:plugin Hope that helps. -chris signature.asc Description: OpenPGP digital signature
Re: Run Applet on Tomcat 5.0.16 -- Where should I to put class files?
Henry, I still get applet not initialized error after I using the archive=http://localhost/PluginApplet.jar; tag in jsp:plugin. What is the actual error? IE should provide a more descriptive error message when the pplet does not load. Try double-clicking on any strange-looking icons in IE's status bar (like the one when you get a JavaScript error). If you're using Sun's VM (with the plug-in), you probably have a Java console running in your system tray. Check that for some kind of log, etc. Do I have to setup classpath environment variable on my Windows system? No. Do you have any external dependencies that are not provided by the Java version you are expecting (like a strange UI library or something)? If so, you'll have to make those JAR files available to the applet by adding them to the archive attribute, separated by commas, like this: jsp:plugin type=applet code=PluginApplet.class archive=http://localhost/PluginApplet.jar,http://localhost/foo.jar; /jsp:plugin signature.asc Description: OpenPGP digital signature
Re: apache 2, tomcat 5, and ROOT application
Dean, I want apache to handle static content and tomcat to handle servlet related stuff (jsp, etc). I am wondering how to specify that in httpd.conf. If I set DocumentRoot to /home/mywebsite/webapps/ROOT/ that will work for http://mywebsite.com/ but it will not find myapp. If I set it to /home/mywebsite/webapps/ it can find http://mywebsite.com/myapp, but then I don't know how it can find the ROOT directory. Here's what you want. For each webapp, you need JkMounts and Aliases. You can use VirtualHosts if you have different server names, but this works, too. # Here's the root stuff DocumentRoot/path/to/webapps/ROOT JkMount /*.jsp root-worker JkMount /whatever root-worker # Here's the stuff for 'app1' Alias /app1 /path/to/webapps/app1 JkMount /app1/*.jsp app1-worker JkMount /app1/whatever app1-worder You can do this for each of your applications. Just alias the base URL to the webapp's directory and add JkMounts. -chris signature.asc Description: OpenPGP digital signature
Displaying a 'down' page with mod_* / Apache
All, I was unable to find anything like this on the mod_* howtos or in the archives. Sorry if I missed somethign blatently obvious. I would like to display a 'down' page when tomcat is not running and the user hits a URL that gets mapped to a worker. Do any of the existing connectors do this sort of thing? Ideally, it would be something like this: JkNoConnectionPage /sorry-app-down-for-maintenance.html ...or something like that. I don't want to just catch error 500 in Apache and use the down page for that, since 500 can occur for any number of reasons -- not just Tomcat being unavailable. I'm currently using mod_jk 1.2.5, but will consider switching to another mod if one supports this kind of thing. Thanks, -chris signature.asc Description: OpenPGP digital signature
Re: verbose gc
Mohammed, 1) [Full GC 34521K-15504K(38328K), 0.5953540 secs] [GC 27533K-16335K(38328K), 0.0396336 secs] 2) what does the above statements got in catalina.out means, is garbage collection active? Yes, garbale collection is active. When the garbage collector runs, it will print a line out to stdout (or stderr?) which ends up in your catalina.out. If it runs a 'full' GC (where it takes a lot of time and cleans up a lot of stuff), then it'll say Full GC, like the first example. Otherwise, it's doing an incremental GC and not hitting everything, like the second example. In a healthy syste, you should see lots of regular GCs, and occationally see a full GC. The two big numbers indicate the size of the active heap (the total size of 'live' objects before and after the GC). So, in the first one, the heap was reduced from 34521k of active objects to 15505k of active objects. The number in parens () is the size of the non-permanent heap, which is the total heap minus one of the surviver spaces. The last number is the wall-clock time that the GC was active -- it tools half a second to so a full GC and about 1/20 of a second to do the incremental GC. Please see http://java.sun.com/docs/hotspot/gc1.4.2/#2.2.%20Measurement|outline for more information. When i execute top command on solaris and see the memory it shows size and resisdent memory , what is the diference between the two. I think this might mean different things to different OSs, but generally the size is the total size of executable + code + heap for that process -- everything it needs to run. The 'resident' is usually the amount of 'size' that's actually in physical RAM and not swapped out to disk by the OS. 3) I see that gc statements are logged to catalina.out but the memory usage keeps increasing, does it mean there is problem with garbage collection, When you say 'memory usage', do you mean from top? Right. Java doesn't return memory to the OS once it requests it -- there's generally no need to do that. Even though Java's heap has more 'space' left over in it, it only has space left over for more Java objects. You should not be comparing the Java heap and GCs to what you read out of 'top'. You should use Runtime.freeMemory (the available heap size), Runtime.totalMemory (the total heap size + internal VM memory), and Runtime.maxMemory (the maximum amount of memory that the VM will try to use). 4) can we get more information like how many obejcts created / objects destroyed during a gc run, what options to pass to get that information You need a profiler to do heap inspections, which has been discussed many times on this list. I'm not sure you can find out exactly what the GC is doing... you might need a very nice profiler for that. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: [OT] Apache and Tomcat together
Mike, Apache is faster for static content. Don't get Yoav started... :) I have gotton him started on this, before. It's funny, though... last time I looked at the Tomcat Connector FAQ, it actually says flat out Apache is faster than Tomcat at serving static content: http://jakarta.apache.org/tomcat/faq/connectors.html. It does qualify that statement by saying that unless you're talking about high traffic, then it doesn't really matter. However, they *do* make that claim. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: problem with arabic in multilanguage jsp
Thomas, Until now, our solution was indeed to put META HTTP-EQUIV=Content-Type CONTENT=text/html;charset=windows-1256 / into the html code. ( Tomcat 4.1.24) If the browser finds a content type in the HTTP header, it ignores the HTML header. What? The whole point of the META tag is actually to *trump* the HTTP header. That's the whole reason that the META HTTP-EQUIV tag exists. For the browser to ignore what you put in there pretty much breaks the rules, right? -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Signal 11 crash and mod_jk, mod_jk2, apache2
All, With RH8, I assume that at the very least you'll need to set LD_KERNEL_ASSUME. I would consider that the first course of action, and likely would not need to do anything else. I could see hyperthreading a problem if the kernel didn't support it very well. You could try the latest 2.4.x kernel. If the LD_ASSUME_KERNEL doesn't help, try disabling both SMP and hyperthreading at the same time. I think that's your next most likely fix. Just a note: I've had big, beefy servers die with SIG11 on Linux before (or course, that was back when a dual athlon 1GHz was considered 'beefy' :). Anyway, we tried everything, including hiring BEA consultants for a bazillion dollars per hour to help us tune both Weblogic and our VM. It turned out to be bad hardware. We had six identical machines and two of 'em kept crapping out. They just sucked. The only solution was to send them back to the manufacturer and ask for more. It turns out that not only did those two (production!) machines suck, but two QA machines and one dev machine (all the same) sucked, too. They all died when we put them under high load. They seemed to do okay under development load (about zero). I'll never buy a machine from Penguin Computing again. Just wanted to mention that sometimes it's not the software's fault. Do you have other similar machines that you can try this on? -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: How can I count the number of active requests/servlets for Tomcat?
Tom,
I'm running a web service using Axis with Tomcat. How can I count
the number of active requests?
I think you want to use a Filter for your requests:
public class ConnectionCounter
implements javax.servlet.Filter
{
private int _activeConnections = 0;
public void init(FilterConfig fc) { }
public void destroy() { }
public doFilter(ServletRequest req,
ServletResponse rsp,
FilterChain fc)
throws ServletException, IOException
{
synchronized(this) {
++_activeConnections;
}
fc.doFilter();
synchronized(this) {
--_activeConnections;
}
}
public int getActiveConnectionCount()
{
// Need not be synchronized, int access is atomic
return _activeConnections;
}
}
Of course, you'll need to get a reference to your filter in order to
read the connection count, but that is an exercise left to the reader :)
I used the HttpListener interace but the sessionDestroyed method
doesn't seem to get called until 5 minutes after the servlet ends.
Ouch. You have web services that use sessions? You should probably turn
off http sessions for those...
-chris
-
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
Re: org/apache/tomcat/logging/Logger
Jake, I got this error in one tomcat application java.lang.NoClassDefFoundError: org/apache/tomcat/logging/Logger It looks like you're missing the log4j library for that particular application. Where you do you see anything about Log4j here? Sorry; complete brain fart, here. Yeah, this has nothing to do with log4j. I was unable to find that class in the JARs that I have in my copy of Tomcat (4.1.29). -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Sessions and clients
Ralph, Is there any way to have the second servlet have access to the same session data that the first one has? Just include the session id in the request. (As query parameter or as form field) This will work as long as you are submitting the request to the same server; if it's going to another server, you'll have to make sure that you use some kind of session replication. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: multiple instance mode on Windows
Venk, Does running multiple instances of tomcat on the same server running on Windows Server 2000, involve running a seperate JVM for each instance? I would think that one of the only reasons for running separate Tomcat instances on the same machine would be for separation of execution environment. In that case, does it even make sense to run them in the same VM? Why do you want multiple Tomcat instances? On the other hand, you might want to run two different versions of Tomcat in a minimum of memory... Also, when using tomcat in production environment, does directing all System.out statements to tomcat console effect its performance? How can this be avoided? Where would I set the JVM options it run the garbage collector in incremental mode? You want to set the CATALINA_OPTS environment variable to include JVM options in your startup script. If you're running Tomcat as a service (which is probably true in production), you'll need to look into how the service actually starts itself (Sorry, I have very little win32/Tomcat experience). -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: problem with arabic in multilanguage jsp
Thomas, we are running a tomcat instance that uses common jsps for all languages. If no encoding is set in the http-header tomcat 4.1.29 sets iso-8859-1 by default. Arabic users must then configure their browsers themselves (unacceptable). (1) If I set the encoding in the jsp like this: [EMAIL PROTECTED] contentType=text/html;charset=windows-1256% everything is fine. But as I told, our jsps are multilingual. (2) If I set the encoding in the jsp like this: % response.setContentType(text/html;charset=windows-1256); % the results are just question marks. The same thing happens if the arabic text is not hardcoded in the jsp but comes from a varable. Have you tried using: META HTTP-EQUIV=Content-Type CONTENT=text/html;charset=windows-1256 / Does this trigger the browser to display the text properly? If so, you might consider adding this for some languages (like Arabic). -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Memory Leaks
Yoav, With JDK 1.4, there's a runtime parameter to ignore System.gc() calls and that parameter may be on by default in the future. What is this parameter? I've never seen it before... -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Memory Leaks
Yoav, With JDK 1.4, there's a runtime parameter to ignore System.gc() calls and that parameter may be on by default in the future. What is this parameter? I've never seen it before... -XX:+DisableExplicitGC is the parameter, which I believe has been available since JDK 1.4.0. Wow, I didn't know that -XX parameters existed. Where can I get more information on these, if you don't mind? Thanks, -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Vedr.: Re: Memory Leaks
Thomas, sorry for interrupting but if you into optimizing your JVM setting as well as explorings its options I would recommend: Perhaps some of the more savvy developers can give more directions. Thanks for the pointers. Actually, I'm not having any performance problems at the moment :) As always remember the best tests are those carried out by yourself profiling you applications - if I didnt say SOMEONE else would have... I've been running OptimizeIt and Tomcat together and quite a while. You don't need to convince me! -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: org/apache/tomcat/logging/Logger
I got this error in one tomcat application java.lang.NoClassDefFoundError: org/apache/tomcat/logging/Logger It looks like you're missing the log4j library for that particular application. Check the WEB-INF/lib directory for a log4j-looking JAR file. It's also possible that the code was compiled with a later version or log4j, and deployed with an earlier version (the Logger class is somewhat new). -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: mod_jk handling all content (not just dynamic)?
Chris, [Sat Dec 20 13:41:33 2003] [jk_uri_worker_map.c (500)]: Attempting to map URI '/admin/something.html' [Sat Dec 20 13:41:33 2003] [jk_uri_worker_map.c (618)]: jk_uri_worker_map_t::map_uri_to_worker, done without a match Even if I comment out all JkMounts in httpd.conf, this problem occurs. Is it really a problem? It seems that jk should not even be seeing requests for static content. Did you re-start Apache after making that change? Are you sure you're modifying the correct file? Send us a clip of your configuration. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Fwd: Help, only viewing source of my index.jsp page - Apache 2.0.48 Tomcat 4.1.29 mod_jk_1.2.5_2.0.47.dll
Lenny, My problem is that when I enter www.mysite.com the source from my index.jsp is displayed. This is almost always a misconfiguration of JkMount directives. It appears that the virtualhost setup in my http.conf is finding the mysite directory in Tomcat, but for some reason cannot display the index page as it should. I have index.jsp setup in my web.xml as my welcome page. VirtualHost www.mysite.com ServerName www.mysite.com # Static files DocumentRoot D:/java/Tomcat41/webapps/mysite Alias /mscarloans D:/java/Tomcat41/webapps/mysite JkMount /mscarloans/*.jsp ajp13 This line maps all JSP files in the /mscarloans/ tree to be sent to Tomcat for processing. There doesn't appear to be any other JkMount for this VirtualHost. I think you want thic changed to something like: JkMount /*.jsp ajp13 Location /mysite/WEB-INF/* AllowOverride None deny from all /Location Good call! Lots of people forget about this! Directory D:/java/Tomcat41/webapps/mysite/WEB-INF/ AllowOverride None deny from all /Directory You should generally do the Directory one. That makes it so that, no matter what strange location and URL games are played, any request that ultimately points to files in this directory are denied. With the Directory config, you don't even need the Location config. Hope that helps, -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: How can I compile Mod_jdk2 ?
William, I use the jakarta-tomcat-connectors-jk2-2.0.2-src and try this: ./configure --with-apxs2=/usr/local/apache2/bin/apxs \ --with-apache2=/usr/local/apache2\ --with-apach2-include=/usr/local/apache2/include\ --with-apache2-lib=/usr/local/apache2/lib \ then in the end of configure process, I received a message: no apxs given building connector for apche-2.0 configure: error: can't locate /usr/local/apache2 Hmmm... do you really need all those options? I would think that: ./configure --with-apxs2=/usr/local/apache2/bin/apxs would be sufficient. Let's see... $ tar xzf jakarta-tomcat-connectors-jk2-2.0.2-src.tar.gz $ cd jakarta-tomcat-connectors-jk2-2.0.2-src/jk/native2 $ ./configure --with-apxs2=/usr/local/apache2/bin/apxs [completed successfully] $ make [completed successfully] $ ls ../build/jk2/apache2/*.so ../build/jk2/apache2/jkjni.so ../build/jk2/apache2/mod_jk2.so $ Looks good to me. I also built my Apache 2.0.48 from source. The code, etc, is in /usr/local/httpd-2.0.48, and not in /usr/local/apache. It's funny that the build instructions in the jk directory say Just type ant!, and it fails miserably. Too bad. I like the whole configure/make/make install. It tends to work well. Let me know if you are still having problems compiling... -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Convert Text to Image [OT]
Andoni, Does anyone know of a tool for Converting Text to images so that I can still have image buttons on my site while having it localised? PS: I have taken a look at AcmeGif I am looking to see if there is something better though. Or tutorial help on AcmeGif. Back when I was working for a client on a web-based website builder (don't ask...), we used a tool called Macromedia Generator would could do tons of sexy things like placing text on top of an existing image, with other text effects, too. We paid like a million bucks a day for the licence or something, and we never used the tool to it's fullest extent. What a shame. Anyhow, that tool exists, and it'll do *everything* you want. I guess it depends on the amount of money you want to spend. I always wanted to do a minimal all-Java re-write of the stuff we needed. It would be faster, cheaper, and easier to manage. (MM Generator only ran on WinNT in those days... all those Sun boxes and a single WinNT box for this tool). If I were you, I'd whip up something quick and dirty around the following core: BufferedImage bi = new BufferedImage(-- sizes --); bi.drawImage(0,0,, etc... if you want to have an image underneath); bi.drawString(center coords, messages.getString(hello.world)); Then, either use AcmeGif or Java's JAI package or the always-there-but-never-supported com.sun.image.codec.jpeg package: http://java.sun.com/j2se/1.4.2/docs/guide/2d/api-jpeg/overview-summary.html Good luck, -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: How can I compile Mod_jdk2 ?
William, I trying to compile the mod_jdk from the source version jakarta-tomcat-connectors-jk2-2.0.2 but everytime an error happens. Please provide more information. The error message itself, perhaps. Does anybody know a place where I can get a binary for linux, to connect my apache 2.0.48 and Tomcat 5.0.16 ? This does not exist. Binary versions are not available from Jakarta for Linux. It's because you need to compile it against your own apache version and libc version. Sure, you *can* get one from someone else, but it shouldn't be a big deal to compile it. Or maybe send-me the module by e-mail That's a dangerous request. Someone could send you a trojan... I using Fedora Core box and J2sdk 1.4.2 If this is impossible, maybe someone just has passed for the same troubles ?I have now, and help me with something. Check the archives for your specific error message. We've beat this to death like a million times on this list. If your error mesage is not found, just followup your own post with the error and we'll see what we can do. Since you're using Fedora, it's very possible that you are missing some required library. Fedora isn't exactly ready for prime time. :) -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Benefits of Apache Server
Yoav, I would be very careful with a claim like Ms. Smoak's, Apache is... much faster serving static resources as that's highly qualitative. Please provide benchmarks or tests that prove your point, because I don't think it's much faster at serving static resources. Retraction! I should have said, Apache is faster at serving static resources than the Tomcat/connector/Apache combination. I've never tested Tomcat alone vs. Apache alone-- Apache was in place here before I started writing webapps. I think we can all agree that this statement is true. Certainly adding Apache to Tomcat is slower than having Tomcat do the work all by itself. I don't have benchmarks, and I've made the statement Apache is faster for static content tthan Tomcat before, and I still believe it to be true. Here's why: When Apache httpd serves a static resource, it pretty much goes through all the processing required to find out where the file actually is, and then dumps the bytes to the response. I'm pretty sure that it does all this without allocating anything on the heap -- most of the object/data structures necesary for the lookup I'm sure already exist, and the buffer for the response is probably fixed, and probably on the stack. For Tomcat to do the same thing, it's got to create a bunch of objects which later need to be garbage-collected. I'm guessing that the Tomcat devs have streamlined the process so that not all of the ServletRequest objects and all that jazz are created every time, but you still have to create a lot of stuff on the heap (including every String used, like the URL, and maybe some headers, etc.). After that (probably after the response has been sent to the client, which is why the numbers are hard to track down), the GC has to run. I think that I can make the blanket statement that explicit memory management is faster than GC'd memory, since the GC actually to do some work to determine if memory can be freed, while the explicit scheme needs no such processing. I am willing to concede that Apache vs. Tomcat in a direct competition for serving up static content on the same hardware will probably result in timing differences so small as to be insignificant to anyone doing reasonable benchmarking. Lastly, if you have your architecture such that you have crappy machines in front of the application servers to serve static content (so that the app servers don't waste time serving static content), Apache will run must better on them since it requires fewer resources to run nicely. For example, Apache can do quite nicely on a 16MB machine as a web server. Tomcat can't really do that since the JVM is such a monster. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: how to change default SMTP port 25 in java mail?
Bopanna, 1. This is not related to Tomcat. Please try a JavaMail list next time. 2. RTFM: http://java.sun.com/products/javamail/javadocs/overview-summary.html Search the page for port. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: help using JDBCRealm - how to relate a realm with a security constraint
Pedro, What i have done: 1. modify the server.xml file to add a Realm tag (Realm className=org.apache.catalina.realm.JDBCRealm debug=99 connectionName=root connectionPassword= connectionURL=jdbc:mysql://myserver:3306/mydatabase driverName=org.gjt.mm.mysql.Driver roleNameCol=role_name userCredCol=user_password userNameCol=user_name userRoleTable=user_roles userTable=users /) Where did you put this declaration? In the Context of your web application? I recall that Realm declarations can appear in multiple places. Make sure that it appears in the Context where you have: 2. modify the web.xml to add a security-constraint tag ( security-constraint web-resource-collection web-resource-nameArea de pruebas de seguridad/web-resource-name url-pattern/cosa.jsp/url-pattern url-pattern/test.jsp/url-pattern http-methodDELETE/http-method http-methodGET/http-method http-methodPOST/http-method http-methodPUT/http-method /web-resource-collection auth-constraint descriptionrol de administrador/description role-nameYYY/role-name /auth-constraint /security-constraint ) This looks good. What is the URL that you expect to be protetced, and is not protected? Also, try removing the http-method settings, just in case that is complicating things. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Image Viewing
All, It would be img src=http://yourserver/yourservlet?param=paramValue; alt=something This still doesn't answer the original interpreted question. I don't think it's possible to display an image on a page and prevent users from browsing to it directly from their browser. The only thing I can think of is to check the REFERER header to see if it came from the page on which you want to display it. That's also not foolproof... -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Cookies.
Abdul, So, I have added the cookies value in servlet, and I can get the cookie value in jsp. When I work with the same browser fine working. But when close and open the new browser window I cant get the cookie values. Setting cookie, res.addCookie(new Cookie(entID,eID)); res.addCookie(new Cookie(lgnId,lgnId)); I checked the javadoc documentation for the Cookie class, and it doesn't seem to mention the default life of a Cookie object once sent to the browser. I'm inclined to think that the default is that it will live as long as the browser session does (especially because of your evidence). Try this: Cookie cookie = new Cookie(entId, eID); cookie.setMaxAge(cookie_life_in_seconds); res.addCookie(cookie); cookie = new Cookie(ognId, ignId); cookies.setMaxAge(cookie_life_in_seconds); res.addCookie(cookie); This is likely to extend the life of your cookies beyond the browser's session. Therefore, they will be sent in requests after the browser is restarted. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Servlets with JDBC connectivity
Todd, SQLUtils.executeQuery(a SQL statement); SQLUtils.executeUpdate(another one); Just out of curiosity, what do these methods return? If the former returns a ResultSet object, then you're in for a world of trouble. The ResultSet will never get closed, or you'll close the connection over which the ResultSet communicates, and then you're hosed, 'cause the calling code can't access the data. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Global Variables
Duncan, and String param = servletContext.getInitParameter(param); Thanks for the replies, but what do I define servletContext as? In every JSP, the ServletContext object is implicitly declared with the identifier application. So, you should just be able to do something like this in your JSP: p my_param = %= application.getInitParameter(my_param) % /p -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Optimizeit remote profiler
Steffen, I have Windows XP as client running the UI and I want to profile my tomcat 4.1.29 server on debian 3.1 sarge running on sun jvm 1.4.2. I have basically the same environment, and mine has worked. The profiler attaches to the server and I do get very few information, but the list of instances stays completely empty. Are you sure that it connects? How are you starting Tomcat? -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Tomcat 4.x non .jsp suffix for includes?
Richard, Does anyone know how to configure tomcat to compile .jnc files in this circumstance? Is there a standard convention for naming jsp include files? Yeah, .jsp :) Seriously, though. Check out CATALINA_HOME/conf/web.xml for how they configure the .jsp file trnaslator/compiler. You could modify the that web.xml file to include the same type of things for .jsp files, or you could modify the deployment descriptor for your application. Unfortunately, modifying your deployment descriptor will make your application non-portable. But, you are already making your .jsp files called something else, anyway, so you'll have to do some sort of re-configuration for every app server you indend to use. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Please Help to compile mod_jk2 on AIX
Jose, These all seem to be warnings, not errors. Does the build succed in creating an object file? Yes object file is created. Okay, I think you're safe ignoring those warnings, then. Are you using apxs to create an Apache DSO? If so, the file apache-2.0/mod_jk.la should exist (or apache-1.3/mod_jk.so for Apache 1.3). You can simply copy it to your Apache modules directory. (Sorry, these are for mod_jk... mod_jk2 is likely to have the filenames slightly different). Basically, you're looking for mod_jk.la for Apache 2.0 and mod_jk.so for Apache 1.3. Good luck, -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Tomcat 4.0.x memory leak (not javac)
All, http://www.apache.org/dist/jakarta/tomcat-4/v4.1.29/RELEASE-NOTES http://nagoya.apache.org/bugzilla/show_bug.cgi?id=20758 Uhhh... this looks like standard operating procedure for Java to me... When you dump a ClassLoader and all of it's object get GC'd, the VM almost never GC's the unused classes. The new ClassLoader re-loads all those classes, again, and you get new versions of those classes loaded into the VM. Even if there are no more instances of those classes, I thought they stuck around... I was not suprised by anything mentioned in that bug. The only thing that's a shame is the fact that it's true, and it does interfere with the ability of an app server to re-load contexts indefinately. :( The -Xnoclassgc option still exists for the 1.4.1 VM, so maybe I'm completely wrong, here. :) -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: SessionListener
Justin, Won't quite do it, JDBCRealm looks for users in a database, I want to connect a user TO a database using their credentials, but the code to do this feat will be quite minimal by comparison. This will make it hard to use a connection pool (which you mentioned that you do/want to do). Is there a particular reason for the database access paranoia? Most apps connect to the db using the same login regardless of the user actually logged in to the application. They use other types of permission checking to see if you can perform some action, instead of relying on thr database for that kind of checking. I absolutely agree that having multiple layers of security is great, but this one may make your application suck really bad, especially if you are using a db like Oracle, where the database connections are anything but lightweight. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: OutOfMemory Exception initializing page context
What's with all the OOM questions this week? I've noticed that topics tend to come in waves. I wonder if there are only like 5 people out there that post to the newsgroup, but they have all kinds of alias and alternate personalities so that the average response is what they use to solve their problems :) Just an amusing thought to make me smile. :) There. I did it. -chris Shapira, Yoav wrote: Howdy, Once you're out of memory, the system is in a bad state and must be restarted. You will see error messages that are at best meaningless and at worst misleading, because they will make you look in the wrong places. You want to fix the memory issue first, by adding more memory (via the -Xmx java command-line parameter) or fixing a memory leak if there's one in your application. Do not assume, for example, that the IllegalStateException you cite has anything to do with the OutOfMemory error: it may be a misleading message thrown by the system after it ran out of memory (even though it shows up before the OutOfMemoryError in the log) while it was thrashing to GC stuff by invalidating your sessions. Yoav Shapira Millennium ChemInformatics -Original Message- From: Rob Wichterman [mailto:[EMAIL PROTECTED] Sent: Tuesday, December 02, 2003 9:57 AM To: [EMAIL PROTECTED] Subject: OutOfMemory Exception initializing page context Hello, I am having major memory problem that I have been fighting through. Has any body ever seen this one. Here are the errors that always come up in my logs. 1. java.lang.IllegalStateException: getAttribute: Session already invalidated 2. JspFactoryImpl: Exception initializing page context java.lang.IllegalStateException: Cannot create a session after the response has been committed 3. JspFactoryImpl: Exception initializing page context java.lang.OutOfMemoryError Any help at all would be greatly appreciated. Thanks, Rob Wichterman This e-mail, including any attachments, is a confidential business communication, and may contain information that is confidential, proprietary and/or privileged. This e-mail is intended only for the individual(s) to whom it is addressed, and may not be saved, copied, printed, disclosed or used by anyone else. If you are not the(an) intended recipient, please immediately delete this e-mail from your computer system and notify the sender. Thank you. - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED] - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: OutOfMemory Exception initializing page context
Rob, Well I have had my -Xmx set to as high as 1gb and I still receive these errors. The site will stay running for at best 1 day without crashing. I was just hoping one of these could point me in a direction while I work on looking at memory leaks. Sorry for the levity. Try this: Turn on verbose GCs, and make sure you're sending stdout to a log file (Tomcat does this by default, I think). Empty the stdout log file, restart Tomcat and then let it run out of memory. Take a look at the GC output (they are prettty obvious) and see what's happening to memory. I've done this in the past, and I even wrote a (long lost) Perl script to parse the log file and generate X-Y coordinates that I could plot using Excel. We could plainly see that there was a memory leak of some kind because the graph kept going up and up. It does take a long time to find memory leaks but there are a couple of things you can check. First of all, what JSK are you using? Check some of the following: - Are you starting any of your own threads? If so, are they dying when the context reloads? - Are you reloading the context a lot? - Do any of your container objects like hashtables or whatever keep references to things like the ServletContext, etc.? - Do you regularly remove old objects from users' sessions? - Are you doing any XSLT? - Are you using a database connection pool or are you creating new connections for each user? - Do you keep any persistent connections like sockets or anything open? -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: load-on-startup and multiple coyotes
Steffen, As soon as I start using 2 coyote connectors together with load-on-startup tomcat initializes the database pools twice. Interestingly it keeps initializing things twice even if I add a third coyote. In about two minutes, Yoav Shapira is going to tell you this: Don't use a servlet to initialize your stuff. Instead, use a ContextListener. IT's cleaner, more flexible, and only gets loaded when it should get loaded. I used to have InitServlet-style initialization, but now I use ContextListeners. They're pretty easy: create a new class that implements javax.servlet.ServletContextListener. Once you move your code from your old InitServlet, you have to install it into the web.xml file like this: listener listener-classfully.qualified.InitListener/listener-class /listener Put this section right before any servlet definitions but after any filter and filter-mapping sections you have. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Tag Handler pool eating up Memory (and enablePooling is set to false)
Neil, This is strange. When I created a heap dump from my server just now, it does not have any instances of TagHandlerPool in the heap dump. That is what I expected before, but I still had them in my last dump. That is weird. However, you still have a bunch of tag handler instances, I see... Doing a quick grep on my summary results from the new heap dump, I get this result: 60256,1076,com/slsideas/pagegen/tags/SetPropertyTag 31296,652,com/slsideas/pagegen/tags/ValueTag The first column is the total bytes held by the instance inself (not including references), the second column is the number of instances that were present in the heap dump, and the last column is the type of the object. So you have 60k in SetPropertyTag objects? If you use them a lot (without pooling), then you'll get a lot of them on the heap at any given time. Can you observe that the number keeps going up, and never goes down? I find it very hard to believe that we have over 2500 active instances of our tags. This seems to imply they are not being garbage collected. Not necessarily. It might just be that they are taking a while to get GC'd. Again, what do you observe over time? I don't belive the StringBuffer is the cause of my problems. Okay, that's good. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Can Tomcat really run cgi scripts?
Ken, But now I'm stuck with this error: java.io.IOException: CreateProcess: perl C:\Tomcat4112\webapps\examples\WEB-INF\cgi\exp.cgi This is likely to be due to the lack of PATH information available to the script itself. Try using #!/usr/bin/perl in your script instead of #!perl I've heard some complains about sub-process permissions, but my initial reaction is that they are errant complaints. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: OutOfMemory Exception initializing page context
Rob, I consulted with the developers and they said we don't use or do anything that you suggested except we do use tomcat's connection pooling for oracle. Okay, that's good. Sometimes, people don't use connection pooling or do it improperly, and leave Connection objects lying around. SQL Connection objects do not clean themselves up nicely, if they even clean themselves up... We are running Sun j2sdk1.4.2_02 with the Concurrent Mark Sweep GC with -Xmx512m. I currently use the verbose logging for GC times. The odd thing is that it does not seem to progressively go up but rather spike up with high GC times and the full GC will not clean anything out. Hmm... that's a little odd. You say you get huge, sudden spikes from which the GC cannot recover? That sounds like bad news. Do you have any data on which pieces of code (servlets, JSPs, actions) are being executed when this happens? Consider adding a request logger filter to see what might be causing the spike... -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: OutOfMemoryError
Trenton, Are you using the context reloading feature? reloadable=true or using the tomcat manager app will cause this to happen eventually. Really? Prove that please... Actually, I can't prove it in his instance, but I have proved it time and time again on ours. I'm using 4.0.4 right now though so I haven't verified 4.1.x. But, I heard that this is still the case in 4.1.x. To reproduce this problem just reload a context over and over. Depending on how much memory it uses, it will start throwing OutOfMemory exceptions long before the -mx JVM parameter value is ever reached. eg. I do a ps on linux and it shows it as using like 140M or something like that and it already starts throwing out of memory exceptions. Aah. I think I see a part of the problem. First of all, the amount of memory shown by 'ps' is completely irrelevant, except that it shows how much memory that the OS has allocated to the Java process. This is allowed to be more than the max heap size set for Java. The heap is not related to the amount of memory the OS allocates, except that the OS report will always be larger than the heap report from Java. You can get an OOM even if you have the heap set to 1GB and yet the process only has 50MB used. This is because OOMs usually happen *while the GC is doing its work*. I'm not sure if there's another thread that increases the actual size of the heap for you, but if so, I'm sure it's related to the GC anyway. So, moving on, what usually happens is that a low memory condition triggers a GC (usually a full GC). When the GC runs, it needs some memory to work with. If it cannot allocate memory for itself, it will die with an OOM. If the heap hasn't increased, yet, then the GC cannot allocate more objects on the heap to do it's thing. That can explain why you get an OOM when you are way under the heap size in the output of ps. With our applications that we're running, we can usually get away with roughly 5-10 context reloads. Do you have any 3rd party containers or such that live outside of the context (their JARs would be in CATALINA_HOME/server/lib or CATALINA_HOME/common/lib -- someone please confirm that the second of these two does *not* get discarded along with the rest of the context upon reload). If you are using 3rd party containers like this, they will not dump their contents when you re-load the servlet context. If you have static singleton instances of these, then you'll likely fill up memory all the time. This could also be related to the Java doesn't GC java.lang.Class objects thread that's going around this week, too. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
Re: Tomcat reload / classloader / connection pool
Marcel, The problems with Tomcat occur upon reload of the web app. The classloader reloads the whole app and all my DB connections get opened again. This wouldn't cause much pain if the existing connections were garbage collected i.e. finalized. This doesn't happen and after a couple of reloads the DB's max connections limit is reached... Where do the class files for your DB connection pool live? Is there a reason to use your own home-brewed CP other than bragging rights? :) First of all, why is this so? This is likely because of the way you use the singleton. When you have a singleton (I use quotes because it's probably not a real singleton otherwise we would not be having this discussion) the static data is associated with the java.lang.Class object that gets created when your class is loaded. When the re-load occurs, that static data sticks around. The new classloader used for the new context loads a new copy of the singleton and they pile up over time. You need to shutdown the connection pool before the context dies. Second of all, how can I prevent this? Somehow listen for reloads and react appropriately? Yes. Consider writing a ServletContextListener and closing the pool before the context goes down. It will be run when the new context comes up, too. Check the documentation for javax.servlet.ServletContextListener Instead of deploying the framework JAR in the web app's WEB-INF/lib directory I could place somewhere in the regular CLASSPATH, but then all web apps would have access to the same ConnectionFactory :-(( Yes, and the problem would get worse. -chris - To unsubscribe, e-mail: [EMAIL PROTECTED] For additional commands, e-mail: [EMAIL PROTECTED]
