row of dashes, and instead check for the dashes, and then some additional
critera.
Here's the complete text of my message:
I've been having a lot of problems with the JRun connector to Apache of
late. I'm running CFMX 6.1, Apache 2.0.47, and RedHat Linux 8.0 on a dual
Xeon with a gig of RAM. Aside from a couple ssh sessions, there is nothing
else running on it.
I'm using http_load against the server at 85 requests per second for a
collection pages and static files. In general, everything seems normal from
the outside (though the log is filling up). However, occasionally Apache
locks up, spikes to 100% of one CPU, and hamstrings CF. It seems to be tied
to the listener thread / segfault errors seen below. Apache is just
spinning it's wheels under that load, no work at all, so I strongly suspect
it's the connector that's having issues.
The apache error_log is full of errors I don't really understand from the
JRun connector:
<start_of_section>
[Thu Oct 16 18:26:03 2003] [notice] jrApache[31627] startSync - detected
broken mutex!
[Thu Oct 16 18:26:05 2003] [notice] jrApache[14476] startSync - detected
broken mutex!
[Thu Oct 16 18:27:57 2003] [notice] jrApache[49555] startSync - detected
broken mutex!
[Thu Oct 16 18:28:32 2003] [notice] jrApache[44702] startSync - detected
broken mutex!
[Thu Oct 16 18:28:34 2003] [notice] jrApache[41118] handleJrppRequest: bad
command: 200 <- [10]
[Thu Oct 16 18:28:34 2003] [notice] jrApache[41118] returning error page for
JRun connector protocol out of sync
[Thu Oct 16 18:28:34 2003] [notice] jrApache[42142] returning error page for
9 Bad file descriptor
[Thu Oct 16 18:28:38 2003] [crit] the listener thread didn't exit
[Thu Oct 16 18:28:39 2003] [notice] child pid 9455 exit signal Segmentation
fault (11)
[Thu Oct 16 18:28:40 2003] [crit] the listener thread didn't exit
<end_of_section>
CF's server.log is devoid of errors, except for logging slow requests. The
cfserver.log contains a slew of messages that look like this:
<start_of_section>
10/16 03:09:40 error Error while reading header S_BROWSER
java.net.SocketTimeoutException: Read timed out
at java.net.SocketInputStream.socketRead0(Native Method)
at java.net.SocketInputStream.read(SocketInputStream.java:129)
at java.io.BufferedInputStream.fill(BufferedInputStream.java:183)
at java.io.BufferedInputStream.read1(BufferedInputStream.java:222)
at java.io.BufferedInputStream.read(BufferedInputStream.java:277)
at jrun.servlet.jrpp.ProxyEndpoint.readFully(ProxyEndpoint.java:530)
at jrun.servlet.jrpp.ProxyEndpoint.readFully(ProxyEndpoint.java:522)
at jrun.servlet.jrpp.ProxyEndpoint.readInt(ProxyEndpoint.java:540)
at
jrun.servlet.jrpp.ProxyEndpoint.readString(ProxyEndpoint.java:569)
at jrun.servlet.jrpp.ProxyEndpoint.getHeader(ProxyEndpoint.java:696)
at
javax.servlet.http.HttpServletRequestWrapper.getHeader(HttpServletReq
uestWrapper.java:124)
at
javax.servlet.http.HttpServletRequestWrapper.getHeader(HttpServletReq
uestWrapper.java:124)
at coldfusion.runtime.CgiScope.resolve(CgiScope.java:123)
at coldfusion.runtime.CgiScope.containsName(CgiScope.java:265)
at coldfusion.runtime.CgiScope.search(CgiScope.java:298)
... this continues ...
<end_of_section>
They appear in groups where the only difference is that S_BROWSER on the
first line roles through a bunch of different values. They don't seem to be
tied to the apache errors, and I've seen them in the logs numerous other
times as well.
Killing off apache seems to fix the problem, though it is very slow to exit
(20-30 seconds). CF seems largely unaffected by the whole process except
that it can't handle the requests, because Apache isn't doing it's share of
the work.
I've Googled numerous times, read through all the relevant technotes I could
find on MM.com, and nothing seems to have improved the situation. Memory
usage for both CF is similar to it's unloaded state after filling it's
template cache (which is larger than the total number of templates on the
server). Response times are slightly more than unloaded (say 40-70%
higher). Apache is also fairly stable with mem consumption, even after a
segfault. Like I said, it's just sitting there, using about 10% of one CPU
at full load.
Any help, thoughts, ideas would be much appreciated.
thanks,
barneyb
-----Original Message-----
From: Barney Boisvert [mailto:[EMAIL PROTECTED]
Sent: Thursday, October 16, 2003 6:49 PM
To: CF-Talk
Subject: Apache issues with CFMX 6.1
I've been having a lot of problems with the JRun connector to Apache of
late. I'm running CFMX 6.1, Apache 2.0.47, and RedHat Linux 8.0 on a dual
Xeon with a gig of RAM. Aside from a couple ssh sessions, there is nothing
else running on it.
I'm using http_load against the server at 85 requests per second for a
collection pages and static files. In general, everything seems normal from
the outside (though the log is filling up). However, occasionally Apache
locks up, spikes to 100% of one CPU, and hamstrings CF. It seems to be tied
to the listener thread / segfault errors seen below. Apache is just
spinning it's wheels under that load, no work at all, so I strongly suspect
it's the connector that's having issues.
The apache error_log is full of errors I don't really understand from the
JRun connector:
[Todays Threads] [This Message] [Subscription] [Fast Unsubscribe] [User Settings]
