Re: Forcing hb-standby with heartbeat when local process in not recoverable

[Martin Pala]

Yes, something like that could be useful. Optionaly it could be possible 
to add the possibility to override the default restart action on the 
process nonexistence event and allow to stack the rules, such as:

--8<--
 check process apache ...
  start program = "/etc/init.d/httpd start"
  stop program = "/etc/init.d/httpd stop"
  if failed existence for 1 cycle then restart
  if failed existence for 3 cycles then exec 
"/usr/lib/heartbeat/hb_standby"
--8<--

The fourth line equals to the implicit restart on process failure and 
most probebly may be omited. The fifth line allows to set the watermark 
- in the case that the process doesn't exist for 5 cycles, then make the 
node standby.


Martin



address@hidden wrote:
> Hello,
>
> I have a question regarding the usage of monit and heartbeat.
> The monit-readme contains the following statements :
>
>  ... "With this sequence, there should be no error-case, which is
> not handled either by heartbeat or by monit. For example, if monit dies,
> initd restarts it. If heartbeat dies, monit restarts it. If the node
> dies, the heartbeat instance on the other node detects it and restart
> the services there."
>
> But what happens, if a service ( for example Apache ) can't be restarted
> locally by monit, because something went wrong with configuration files
> etc ?
> Then the service would not be accessible any more, because the virtual
> address is
> still on the failed node. Wouldn't it be a good idea to have something
> like:
>
> IF NUMBER RESTART NUMBER CYCLE(S) THEN ACTION, where ACTION could be
> EXEC and not only TIMEOUT ?
>
> Then it would be possible to exec "hb_standby" if Apache could not be
> restarted for 3 times within 3 cycles
>
>  check process apache
>            with pidfile "/usr/local/apache/logs/httpd.pid"
>            start program = "/etc/init.d/httpd start"
>            stop program = "/etc/init.d/httpd stop"
>            if 3 restarts within 3 cycles then exec
> "/usr/lib/heartbeat/hb_standby"
>
>
> Or is there a better way to achieve this ?
>
> I know that heartbeat 2.x can do ressource monitoring, but I have to
> rely on heartbeat 1.2.3 for several reasons.
>
> Thanks for your help in advance
>
> Best regards
>
> Tim