Re: [monit] monit starts a process twice

[Martin Pala]

Hi,

thanks for data, the problem is, that monit enables the monitoring too 
early - as soon as it creates the wait_start thread. If the process is 
starting slowly, the monitoring detects that it is not running but it 
doesn't know that the start is pending already.

Here is patch which should fix the problem.


Thanks,
Martin





Nick Upson wrote:
> Here is an extract from the log, with additonal comments as I *may*
> have an idea of the problem
>
> Mon Feb  4 08:28:07 MST 2008 restart bs1
> # this is me issueing the command for the restart
>
> [MST Feb  4 08:28:07] info     : restart service 'bs1' on user request
> [MST Feb  4 08:28:07] info     : monit daemon at 10762 awakened
> [MST Feb  4 08:28:07] info     : Awakened by User defined signal 1
> [MST Feb  4 08:28:07] info     : 'bs1' trying to restart
> [MST Feb  4 08:28:07] debug    : Monitoring disabled -- service bs1
> [MST Feb  4 08:28:07] info     : 'bs1' stop: /opt/unb/bin/bs.sh
> # everything as expected so far
>
> [MST Feb  4 08:28:08] debug    : 'bs1' Error testing process id
> [10793] -- No such process
> [MST Feb  4 08:28:08] debug    : 'bs1' Error testing process id
> [10793] -- No such process
> [MST Feb  4 08:28:08] debug    : 'bs1' Error testing process id
> [10793] -- No such process
> # now why is monit complaining about the process it says it has
> stopped monitoring
>
>
> [MST Feb  4 08:28:08] info     : 'bs1' start: /opt/unb/bin/bs.sh
> # this does work and did start it running as shown by ps output
>
>
> [MST Feb  4 08:28:08] debug    : 'bs1' Error testing process id
> [10793] -- No such process
> # and by this point the pid in the pid file should have the new pid in it
>
> [MST Feb  4 08:28:08] debug    : Monitoring enabled -- service bs1
>
> .
> .
> .
> .
> [MST Feb  4 08:28:09] debug    : monit: pidfile '/var/run/bs1.pid'
> does not exist
> [MST Feb  4 08:28:10] debug    : monit: pidfile '/var/run/bs1.pid'
> does not exist
> [MST Feb  4 08:28:11] debug    : monit: pidfile '/var/run/bs1.pid'
> does not exist
> # continues this until the next scheduled check on processes
> .
> .
> .
> [MST Feb  4 08:30:08] debug    : monit: pidfile '/var/run/bs1.pid'
> does not exist
> [MST Feb  4 08:30:08] error    : 'bs1' process is not running
> [MST Feb  4 08:30:08] info     : 'bs1' trying to restart
> [MST Feb  4 08:30:08] debug    : Monitoring disabled -- service bs1
> [MST Feb  4 08:30:08] debug    : monit: pidfile '/var/run/bs1.pid'
> does not exist
> [MST Feb  4 08:30:08] debug    : monit: pidfile '/var/run/bs1.pid'
> does not exist
> [MST Feb  4 08:30:08] info     : 'bs1' start: /opt/unb/bin/bs.sh
> # so here we go starting the second copy of the same process running
>
> [MST Feb  4 08:30:08] debug    : monit: pidfile '/var/run/bs1.pid'
> does not exist
> [MST Feb  4 08:30:08] debug    : Monitoring enabled -- service bs1
>
> # and from here on monit is happy about the second copy but the first
> one is still there as well
>
>
> --
> To unsubscribe:
> http://lists.nongnu.org/mailman/listinfo/monit-general
Index: control.c
===================================================================
RCS file: /sources/monit/monit/control.c,v
retrieving revision 1.108
diff -u -r1.108 control.c
--- control.c   8 Mar 2008 01:08:39 -0000       1.108
+++ control.c   13 Mar 2008 21:47:33 -0000
@@ -309,9 +309,10 @@
        LogError("Warning: Failed to create the start controller thread. "
            "Thread error -- %s.\n", strerror(status));
       }
+    } else {
+      Util_monitorSet(s);
     }
   }
-  Util_monitorSet(s);
 }
 
 
@@ -480,6 +481,8 @@
     Run.wait_start--;
   END_LOCK;
 
+  Util_monitorSet(s);
+
   return NULL;
   
 }