[Resin-interest] thread-pool configuration in Resin 3.0.19
Title: thread-pool configuration in Resin 3.0.19 We have recently seen some instances where Resin 3.0.19 running with mod_caucho seems to stop answering requests. It's still running, because I've turned on debug logging, and when this happens it continues to write entries about cleaning up the expired sessions, but no new requests get fulfilled. Stopping and restarting resin clears up the issue. Our suspicion is that resin may be running out of threads, so I've been trying to increase the threads by adding a directive to the configuration file >> 300 10 << The documentation suggests in places that this should be a child of and in other places it seems to be listed as a child of - which is correct? I have tried to add it at both levels, and I get a strange error when starting up resin which says that the web-app root directory is the same as resin.home. The relevant lines from the log follow. This error does not appear when I remove the entries. >> [2006/11/08 16:08:48.032] web-app root directory should not be the same as resin.home [2006/11/08 16:08:48.032] /export/home/mltv/resin [2006/11/08 16:08:48.118] WebApp[] starting [2006/11/08 16:08:49.655] Resin restarting due to configuration change [2006/11/08 16:08:49.655] Server[f] stopping [2006/11/08 16:08:49.710] Received close event [2006/11/08 16:17:28.938] Server[f] starting [2006/11/08 16:17:28.940] [2006/11/08 16:17:28.943] SunOS 5.10 sparc [2006/11/08 16:17:28.945] Java 1.5.0_06-b05, 32, mixed mode, ISO646-US, en, Sun Microsystems Inc. [2006/11/08 16:17:28.948] resin.home = /export/home/mltv/resin [2006/11/08 16:17:28.950] server.root = /export/home/mltv/resin << Here is our resin configuration file for reference. Ideas or comments are appreciated. >> http://caucho.com/ns/resin" xmlns:resin="http://caucho.com/ns/resin/core"> timestamp="[%m/%d/%y %H:%M:%S.%s] " rollover-period="1D" rollover-count="7"/> timestamp="[%m/%d/%y %H:%M:%S.%s] " rollover-period="1D" rollover-count="7"/> timestamp="[%Y/%m/%d %H:%M:%S.%s] " rollover-period='1h' rollover-count='4'/> timestamp="[%Y/%m/%d %H:%M:%S.%s] " rollover-period='1h' rollover-count='2'> jdbc/mysql jdbc:mysql://10.10.240.20:3306,10.10.240.21:3306/wad_session?autoReconnect=true&failOverReadOnly=false mltv resin 8 1000 60s jdbc/mysql com.caucho.jsp.JspPrecompileListener jsp jspx xtp << Michael Ebeling Systems Administrator MLT Vacations, Inc. ___ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest
Re: [Resin-interest] Different jvm.log files?
> Can I configure Resin to send JVM outputs to another filename then "jvm.log"? > The reason is that I would like to run several Resin servers on the same machine and therefore I want to configure each resin to do a garbage collection verbose to different "jvm.log" files. > We are running Resin 3.0.22 on a Microsoft Windows Server 2003 R2 with Java 1.4.2_12. Yes. This is not difficult to do. I have a similar setup, although I'm running on Solaris, so the syntax may be a little bit different. I wrote a wrapper script which takes a sitecode as an argument, and then runs httpd.sh like this: >> site=$1 echo "START resin [$site]" ./httpd.sh -server a \ -conf ../conf/$site.conf \ -pid ../$site.pid \ -jvm-log ../log/jvm-$site.log \ start << each site uses it's own conf file, it's own pid file, and writes it's own jvm-log. You will obviously have to modify this so that it works properly on Windows, but the basic idea is that the startup script can take an argument which specifies the name of the jvm.log. This has been great for troubleshooting, especially in combination with the "-verbosegc" flag to the JVM. ___ resin-interest mailing list resin-interest@caucho.com http://maillist.caucho.com/mailman/listinfo/resin-interest
Re: [Resin-interest] Resin stops logging if logs are deleted manually
> -Original Message-
> Sent: Tuesday, June 05, 2007 11:54 AM
> Subject: [Resin-interest] Resin stops logging if logs are deleted
manually
>
> We are using Resin-3.0.19. Our access log is configured in resin.conf
as follows
>
> format='%h %l %u %t "%r" %s %b "%{Referer}i" "%{User-Agent}i" Time
of request:<%Ts>'
>rollover-period="1D" archive-format="access-%Y%m%d.log.gz"/>
>
> If a utility like logrotate deletes the logs, resin stops logging
> completely. The server continues to run and serve requests.
Since you are using resin to roll the logs for you, it would probably be
better not to use logrotate to manipulate the same logs. Resin will also
delete the oldest logs for you when it rolls them if you add the
"rollover-count" directive:
This would save a week's worth of logs plus the current one, for
example.
> The same behaviour is observed for :-
> stdout.log & stderr.log (configured at command line)
We have entries in the resin conf file to handle our stderr-log and
stdout-log the same way.
___
resin-interest mailing list
resin-interest@caucho.com
http://maillist.caucho.com/mailman/listinfo/resin-interest
