Re: [monit] Question about multi-host testing

[Martin Pala]

The problem is, that the dependency was designed primarily for critical 
actions (start/stop/restart/monitor/unmonitor), where correct order is 
needed.

The alert-only action doesn't trigger the dependency (action chain) 
since it could be just informative.

For example if you are monitoring the icmp, you can have few error 
levels, such as:

--8<--
check host myrouter with address ...
  if failed icmp type echo for 3 times within 5 cycles then alert
  if failed icmp type echo for 5 cycles then exec 
"/script/to/power-cycle/router"
--8<--

In such case monit sends alert when the network has problems, but is not 
completely dead (part of packets lost) and can recover itself yet. In 
such case this shouldn't disable the monitoring of remote hosts. When 
the error ratio is 100% for 5 cycles (the second icmp line), then it can 
exec for example script to power-cycle the router (networked power 
switch ... point-to-point or on the same ethernet switch to be reachable 
if router is not available).

So, the final solution could be to extend the dependency and make the 
service dependency hard by option even on alert message (to stop 
monitoring the other services).

Workaround could be to define dummy start/stop methods for monitored 
remote hosts and use restart action instead of alert (it sends alert as 
well). Something like:

--8<--
check host myswitch ...
  start program = "/bin/true"
  stop program = "/bin/true"
  if failed icmp type echo for 5 cycles then restart

check host myrouter ...
  start program = "/bin/true"
  stop program = "/bin/true"
  if failed icmp type echo for 5 cycles then restart
  depends on myswitch
--8<--

... not tested, but can work (although the restart action doesn't look 
logical, it can trigger the dependency in this case as well).


Martin


Pablo Iranzo Gómez wrote:
>         List, here is the output from monit running in interactive mode with
> -vv:
>
> From log start:
> -----------------------------------------------------------------------
> Remote Host Name      = ro5000-siNmG20876YFyCu20879
>  Monitoring mode      = active
>  ICMP                 = if failed Echo Request count 1 with timeout 10
> seconds 1 times within 1 cycle(s) then alert else if passed 1 times
> within 1 cycle(s) then alert
>  Alert mail to        = address@hidden
>    Alert on           = All events
>    Alert reminder     = 1 cycles
>
> Remote Host Name      = pos10.5000-siNmG20876YFyCu20879
>  Monitoring mode      = active
>  Depends on Service   = ro5000-siNmG20876YFyCu20879
>  ICMP                 = if failed Echo Request count 1 with timeout 10
> seconds 1 times within 1 cycle(s) then alert else if passed 1 times
> within 1 cycle(s) then alert
>  Alert mail to        = address@hidden
>    Alert on           = All events
>    Alert reminder     = 1 cycles
>
>
> From Log checking:
> -----------------------------------------------------------------------
> 'ro5000-siNmG20876YFyCu20879' icmp ping failed
> 'ro5000-siNmG20876YFyCu20879' failed ICMP test [Echo Request]
> ICMP failed notification is sent to address@hidden
> 'ro5000-siNmG20876YFyCu20879' icmp ping failed, skipping any port
> connection tests
> 'pos10.5000-siNmG20876YFyCu20879' icmp ping failed
> 'pos10.5000-siNmG20876YFyCu20879' failed ICMP test [Echo Request]
> ICMP failed notification is sent to address@hidden
> 'pos10.5000-siNmG20876YFyCu20879' icmp ping failed, skipping any port
> connection tests
>
>
> Config files:
> -----------------------------------------------------------------------
> check host ro5000-siNmG20876YFyCu20879 with address 10.39.16.1
>         if failed ICMP type ECHO count 1 timeout 10 seconds then alert
>         alert address@hidden with reminder on 1 cycle
>
> check host pos10.5000-siNmG20876YFyCu20879 with address 10.39.16.10
>         if failed ICMP type ECHO count 1 timeout 10 seconds then alert
>         alert address@hidden with reminder on 1 cycle
>         depends on ro5000-siNmG20876YFyCu20879
>
>
>
>         Any hint?
>
>         Thanks in advance,
>         Pablo
>
>
> El lun, 29-10-2007 a las 21:51 +0100, Pablo Iranzo Gómez escribió:
> >         Martin,
> >
> > On Mon, 29 Oct 2007, Martin Pala wrote:
> > > Can you run monit in verbose mode (-v option) and send the log? You'll
> > > see in it what happened in more detail.
> >         Sure, will do it tomorrow early in the morning :)
> >
> > > >         If I just put "if failed icmp then alert" monit complains about
> > > > configuration (I'm using monit-4.9-1), so either I'm doing something
> > > > wrong or it's a problem with this verson.
> > > I'm sorry - this was typo (i wrote the example just from memory so, the
> > > "type echo" was missing).
> >         Don't worry, I was just trying just in case I did something wrong
> > :)
> >
> >         Thanks again
> >         Pablo
> >
> >
> > --
> > To unsubscribe:
> > http://lists.nongnu.org/mailman/listinfo/monit-general
> >
> > ------------------------------------------------------------------------
> >
> > --
> > To unsubscribe:
> > http://lists.nongnu.org/mailman/listinfo/monit-general