Re: pidfile/start: wait for X cycles before restart

[Martin Pala]

Hi,

the solution is to restart the service via monit (for example "monit 
restart myservice") ... this can be done easily by modification of the 
logrotate script for the given service.

This approach is natural and can be seen in many different process 
monitors as well - any mangling of the process should be done via the 
process monitor's interface. For example Sun Cluster or SMF behavior is 
the same, other cluster frameworks will behave most probably same way.

Martin

dAniel hAhler wrote:
> Hello,
>
>
> monitoring a service with "pidfile" and having a "start" command
> given, monit will start this process, if it does not exist anymore.
>
>
> I'd like to give a number of cycles for how long monit should wait
> before it restarts the service, similar to the general "[<X>] <Y>
> CYCLES". E.g., monit would wait for 1 cycle before restarting (instead
> of 0).
>
>
> The reason is just in the process of being manually restarted, e.g. by
> logcheck and monit should not conflict with it.
>
>
> I see that a workaround to this would be to no bypass monit in e.g.
> "logrotate" or when manually restarting a service, but think that this
> is only a workaround, no solution.
>
>
> I'm not sure if the optional <X> here would make sense and where to
> add it: maybe to the "start" command?
>
>
> What do you think?