Jun 1, 2006 4:04 AM
thread deadlock? anyway i can help debug?
-
Like (0)
hi,
i''m using xiff/flash(group chat) client side and wildfire on the serverside.
i did something somewhat crazy today, i put the flash chat which connects to a certain room, on every result on my homepage, which gets at least 1 hit per second, making massive connects/disconnects.
last time i checked i had over 300 concurrent users in the chat room.
anyways, the chat room would freeze up a few seconds every 5~6 minutes, which was still ok, but two times today the server would just freeze, and not respond to even
./wildfire stop
./wildfire start
wildfire stop would say wildfire stopped but it didn''t and i had to kill it manually.
it froze "completely" twice today, once even kill didn''t work and i just rebooted(didn''t try kill -9 though)
the server was running only wildfire(LAMP setup but only the flash file being sent through apache and wildfire was in use, all other homepage stuff is in a different server)
anyways i''m thinking a thread deadlock occured, is their anyway i could help debug?
send error.log etc?
not sure how to use debug.log ...
b.t.w. the server load when it froze was exactly 0.0
thanks.
Hi,
there is a KB article on how to obtain a thread dump, kill -3 $PID should be fine for unix systems. There you would see if a deadlock did occur.
LG
i''ll try that.
uh... what does "kb article" mean? :P
i searched google, seems to be a common phrase, but i can''t seem to define it :P
Re: thread deadlock? anyway i can help debug?Different ways to obtain a thread dump of the Java VM[/url] - with KB = (Jivesoftwares) Knowledge Base, you may enter it here http://www.jivesoftware.org/community/index.jspa if you want to take a look at it as it contains some (only
) useful documents.
LG
sorry :P i''m not that familiar with kill -3... but shouldn''t a kill -3 log it in stdout.log or something like threaddump.. in jre directory or log directory?
Re: thread deadlock? anyway i can help debug?# ps -ef|grep bot
bot 7683 1 2 18:10 ? 00:08:21 /opt/wildfire/jre/bin/java -server -Dinstall4j.jvmDir=/opt/wildfire/jre -Dinstall4j.appDir=/opt/wildfire -Dexe4j.moduleName=/opt/wildfire/bin/wildfire -classpath /opt/wildfire/.install4j/i4jruntime.jar:/opt/wildfire/lib/activation.jar:/opt/w ildfire/lib/bouncycastle.jar:/opt/wildfire/lib/commons-el.jar:/opt/wildfire/lib/ hsqldb.jar:/opt/wildfire/lib/hsqldbutil.jar:/opt/wildfire/lib/jasper-compiler.ja r:/opt/wildfire/lib/jasper-runtime.jar:/opt/wildfire/lib/jdic.jar:/opt/wildfire/ lib/jtds.jar:/opt/wildfire/lib/mail.jar:/opt/wildfire/lib/mysql.jar:/opt/wildfir e/lib/postgres.jar:/opt/wildfire/lib/servlet.jar:/opt/wildfire/lib/startup.jar:/ opt/wildfire/lib/wildfire.jar com.install4j.runtime.Launcher start org.jivesoftware.wildfire.starter.ServerStarter false false /opt/wildfire/bin/../logs/stderror.log /opt/wildfire/bin/../logs/stdoutt.log true true false true true 0 0 20 20 Arial 0,0,0 8 500 version 2.6.2 20 40 Arial 0,0,0 8 500 -1 -DwildfireHome=/opt/wildfire -Dwildfire.lib.dir=/opt/wildfire/lib start
root 16692 16424 0 23:57 pts/0 00:00:00 grep bot
Re: thread deadlock? anyway i can help debug?# kill -3 7683
Re: thread deadlock? anyway i can help debug?# pwd
/opt/wildfire/jre/bin
Re: thread deadlock? anyway i can help debug?# ls -al
total 1348
drwxr-xr-x 2 bot users 4096 May 31 21:41 .
drwxr-xr-x 4 bot users 4096 May 31 21:41 ..
-rwxr-xr-x 1 bot users 4153 Apr 21 02:12 ControlPanel
-rwxr-xr-x 1 bot users 64812 Apr 21 02:12 java
-rwxr-xr-x 1 bot users 26366 Apr 21 02:12 java_vm
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 keytool
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 kinit
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 klist
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 ktab
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 orbd
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 pack200
-rwxr-xr-x 1 bot users 72760 Apr 21 02:12 policytool
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 rmid
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 rmiregistry
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 servertool
-rwxr-xr-x 1 bot users 72752 Apr 21 02:12 tnameserv
-rwxr-xr-x 1 bot users 394235 Apr 21 02:12 unpack200
Re: thread deadlock? anyway i can help debug?#
Re: thread deadlock? anyway i can help debug?# pwd
/opt/wildfire/logs
Re: thread deadlock? anyway i can help debug?# ls -la
total 296
drwxr-xr-x 2 bot users 4096 Jun 2 00:33 .
drwxr-xr-x 11 bot users 4096 May 31 21:41 ..
-rw-rr 1 bot users 3063 Jun 1 17:53 admin-console.log
-rw-rr 1 bot users 0 May 31 21:43 debug.log
-rw-rr 1 bot users 96269 May 31 22:29 error_1.log
-rw-rr 1 bot users 162680 Jun 1 20:12 error.log
-rw-rr 1 bot users 3341 Jun 1 18:10 info.log
-rw-rr 1 bot users 271 Jun 1 18:09 stderror.log
-rw-rr 1 bot users 0 Apr 21 02:09 stderr.out
-rw-rr 1 bot users 117 Jun 1 18:10 stdoutt.log
-rw-rr 1 bot users 0 May 31 21:43 warn.log
Re: thread deadlock? anyway i can help debug?#
Re: thread deadlock? anyway i can help debug?# ps -ef|grep bot
bot 7683 1 2 Jun01 ? 00:10:07 /opt/wildfire/jre/bin/java -server -Dinstall4j.jvmDir=/opt/wildfire/jre -Dinstall4j.appDir=/opt/wildfire -Dexe4j.moduleName=/opt/wildfire/bin/wildfire -classpath /opt/wildfire/.install4j/i4jruntime.jar:/opt/wildfire/lib/activation.jar:/opt/w ildfire/lib/bouncycastle.jar:/opt/wildfire/lib/commons-el.jar:/opt/wildfire/lib/ hsqldb.jar:/opt/wildfire/lib/hsqldbutil.jar:/opt/wildfire/lib/jasper-compiler.ja r:/opt/wildfire/lib/jasper-runtime.jar:/opt/wildfire/lib/jdic.jar:/opt/wildfire/ lib/jtds.jar:/opt/wildfire/lib/mail.jar:/opt/wildfire/lib/mysql.jar:/opt/wildfir e/lib/postgres.jar:/opt/wildfire/lib/servlet.jar:/opt/wildfire/lib/startup.jar:/ opt/wildfire/lib/wildfire.jar com.install4j.runtime.Launcher start org.jivesoftware.wildfire.starter.ServerStarter false false /opt/wildfire/bin/../logs/stderror.log /opt/wildfire/bin/../logs/stdoutt.log true true false true true 0 0 20 20 Arial 0,0,0 8 500 version 2.6.2 20 40 Arial 0,0,0 8 500 -1 -DwildfireHome=/opt/wildfire -Dwildfire.lib.dir=/opt/wildfire/lib start
root 18823 16424 0 00:39 pts/0 00:00:00 grep bot
Re: thread deadlock? anyway i can help debug?#
process still alive.
Hi,
it will write to nohup.out in wildfire/bin and it will of course not kill the process, so one can get more than one dump and keep the application running.
LG
thanks, this is very helpful~~
Re: thread deadlock? anyway i can help debug?# pwd
/opt/wildfire/bin
Re: thread deadlock? anyway i can help debug?# ls -al
drwxr-xr-x 3 bot users 4096 Jun 2 00:47 .
drwxr-xr-x 11 bot users 4096 May 31 21:41 ..
drwxr-xr-x 2 bot users 4096 May 31 21:41 extra
-rw------- 1 bot users 272018 Jun 1 23:55 nohup.out
-rwxr-xr-x 1 bot users 7213 Apr 21 02:12 wildfire
Re: thread deadlock? anyway i can help debug?#
file is HUGE... what should my next step be :P
Full thread dump Java HotSpot(TM) Server VM (1.5.0_06-b05 mixed mode):
"Client SR - 11316080" daemon prio=1 tid=0xa2d01888 nid=0x4095 sleeping[0xa7c9d000..0xa7c9df30]
at java.lang.Thread.sleep(Native Method)
.....
272 kb...
Hi,
"grep -i dead nohup.out" - it should display some lines if there is a deadlock, then you can use an editor to locate the threads. And check the content of stderror.log, I hope that you don''t see OutOfMemory errors there.
LG
thanks, i tried it when the chat was in a "few second frozen" mode :P. chat server is responding though, unlike the two times when it seems to "really freeze".(i''ve been running wildfire for months on a low volume site with no problems, but only for 24hrs on my high volume site which is giving me problems)
"grep -i dead nohup.out" -> no dead threads... stderror.log is essentially empty.
i''ll try the kill -3 next time i see the server "really freeze".
right now i have only 150 users concurrently in the chat room, it froze last time after getting over 300 users... so i think i''ll have to wait till peak visiting time again. unless i get rid of the chat room before then because it''s harming the quality of my site
.
Digirave, did you ever get more information on this problem? I''m having something similar.