----------------------------------------------------------------
BEFORE YOU POST, search the faq at <http://java.apache.org/faq/>
WHEN YOU POST, include all relevant version numbers, log files,
and configuration files. Don't make us guess your problem!!!
----------------------------------------------------------------
Did you check out the output from the logs to get more of an idea what's
happening? Did you try increasing the ApJServVMTimeout parameter in the
jserv.conf to see if it had any effect?
> Besides, it kinda bites that manually kicking off the JVM doesn't seem to
> get proper classpath info from the wrapper.classpath= lines.
You need to create the CLASSPATH environment variable in your manual jserv
script, or pass the classpath information to the jvm when you call it with
the "-classpath" option.
Kevin
-----Original Message-----
From: Williams, Murray Todd <[EMAIL PROTECTED]>
To: '[EMAIL PROTECTED]'
<[EMAIL PROTECTED]>
Date: Wednesday, May 10, 2000 5:25 PM
Subject: Many JServ threads being made and dying
----------------------------------------------------------------
BEFORE YOU POST, search the faq at <http://java.apache.org/faq/>
WHEN YOU POST, include all relevant version numbers, log files,
and configuration files. Don't make us guess your problem!!!
----------------------------------------------------------------
I've gone through the FAQ and the last three weeks of mailing list archives,
and I've seen one person mention this problem, but no further discussion.
I'm the (unofficial) maintainer of the RPM packages for Xerces, Xalan and
Cocoon (these RPMs being built on top of the "official" JServ RPM that is
posted on the java.apache.org web site. For the last few months I noticed
something a little odd, but I didn't pay much attention. I've finally
decided to look closer and I'm very concerned.
I tried submitting a bug report on the automated web form, but I've tried on
two days from two different locations (home and work) and I keep getting and
error back telling me that all my form inputs are blank, so I guess that
isn't working. Instead I've copied all the sections down here...
***Environment:***
I have this occuring on two platforms: RedHat Linux 6.1 and 6.2 (Intel) and
on
LinuxPPC (specifically YellowDog) for the PowerPC. I recompiled the .src.rpm
file
on PPC to create the installation rpm. The Java environments include
Blackdown's
JDK 1.2.2RC4 on both Intel and PPC, as well as IBM's new JDK1.3 for Intel.
***Synopsis:***
A bunch of zombie (defunct) jdk theads and constantly being created and
destroyed.
***Full Description:***
Once apache is started (/etc/rc.d/init.d/httpd start) the Jserv module also
starts up the JDK (specified in jserv.properties). If the JDK is running
with
native threads, the problem is most visible. If you do a "ps aux" to examine
the running processes/threads five or six times, you'll see that there is
almost
always a "[java <defunct>]" process with a new pid. I just did two "ps"
statements one minute apart and in that time there were 20 java threads
created and destroyed.
The problem can also been seen by executing "netstat". You will see that
there
are about ten processes which are stuck in the TIME_WAIT state, waiting to
time-out.
For a while I thought the problem only existed with java native threads so I
switched to green threads. Then I discovered that although there wasn't the
new PID creation/destruction visible with "ps aux" but "netstat" was showing
constantly newly created and destroyed IP bindings on port 8007.
If, however, one sets jserv.conf ApJServManual "on" and launches the JDK
manually, the problem does not happen, the threads do not get
created/destroyed
and "netstat" doesn't show about 10 defunct timed-out TIME_WAIT IP
connections.
***How can we repeat this problem?***
Install a fresh RedHat Linux 6.1 or 6.2. Install the ApacheJServ RPM from
the official web site. Restart apache. This may or may not happen with
hand-built
installations. (I haven't tested that option.)
Is this a known problem? I'm actually amazed that I didn't see if on the
FAQ-o-Matic since I imagine a lot of people use the ApacheJServ RPM package.
(And it might also exist on manual Linux installations!)
I'd love to hear some feedback. I wish I knew more about writing daemon
processes, 'cause I'd be willing to write a launcher/monitor/nuker to put
under /etc/rc.d/init.d, but I don't want to get it wrong if I'm offering the
workaround to the outside world. Besides, it kinda bites that manually
kicking off the JVM doesn't seem to get proper classpath info from the
wrapper.classpath= lines.
Sigh.
Murray Todd Williams
Merck & Co., Inc.
--
--------------------------------------------------------------
Please read the FAQ! <http://java.apache.org/faq/>
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Archives and Other: <http://java.apache.org/main/mail.html>
Problems?: [EMAIL PROTECTED]
--
--------------------------------------------------------------
Please read the FAQ! <http://java.apache.org/faq/>
To subscribe: [EMAIL PROTECTED]
To unsubscribe: [EMAIL PROTECTED]
Archives and Other: <http://java.apache.org/main/mail.html>
Problems?: [EMAIL PROTECTED]
