Re: [monit] "failed to stop" messages

[Jan-Henrik Haukeland]

The stop (and start) command must be a program or script. Monit does not start 
an interactive shell to execute the command, but calls exec. To write a 
shell-script as the command you must therefor first start a shell. 

If you change your stop command to start a shell and then enter your "script" 
it will hopefully work:

 stop program  = "/bin/bash -c '/usr/bin/kill -9 `cat 
/shared/pids/thin.8007.pid` && rm -f /shared/pids/thin.8007.pid'"





On 20. feb. 2010, at 18.57, Dylan Stamat wrote:

> Hello!
> 
> I'm using Monit to monitor some processes, and can't seem to get my simple 
> configuration working correctly.
> When my threshold is met, I end up getting sent constant "failed to stop" 
> messages.
> 
> Here is the output in my logs:
> ---------------------------------------------------------------------------------------------------------------------------------------------------------
> monit[4823]: 'thin8007' total mem amount of 205988kB matches resource limit 
> [total mem amount>163840kB]
> monit[4823]: 'thin8007' trying to restart
> monit[4823]: 'thin8007' stop: /usr/bin/kill
> monit[4823]: 'thin8007' failed to stop
> ---------------------------------------------------------------------------------------------------------------------------------------------------------
> 
> Here is my configuration:
> ---------------------------------------------------------------------------------------------------------------------------------------------------------
> set daemon  20
> set logfile syslog facility log_daemon
>   check process thin8007 with pidfile /shared/pids/thin.8007.pid
>   start program = "/usr/bin/thin start -C /etc/thin/application.yml --only 
> 8000"
>   stop program  = "/usr/bin/kill -9 `cat /shared/pids/thin.8007.pid` && rm -f 
> /shared/pids/thin.8007.pid"
>   if totalmem > 160.0 MB for 1 cycles then restart
>   if cpu > 90% for 1 cycles then restart
>   group thin
> ---------------------------------------------------------------------------------------------------------------------------------------------------------
> 
> As you can see, the "stop" directive is a bit of a brute force method.  Prior 
> to using that, I was using the "stop" command
> of the application (thin) I'm trying to monitor.  I ran into a problem when 
> the application wouldn't clean up after itself, and
> it would end up leaving stale pid files around.  So, I decided to SIGKILL the 
> process and clean up the pid manually.
> 
> If I run the stop command manually, the process is killed and the pid file is 
> gone.  However, when it is run through Monit, I
> get the "failed to stop" message.  Monit is run as root on this system, but, 
> it still seems like it could be a permissions issue?
> Is there anyway to get more verbose output in regard to why it "failed to 
> stop"?  Is there anything that Monit could glean from
> the output of the system calls it makes?  I'd be happy to patch if that was a 
> possibility!
> 
> Any suggestions would be welcome!
> Thanks!
> ==
> Dylan
> 
> --
> To unsubscribe:
> http://lists.nongnu.org/mailman/listinfo/monit-general