On 7/23/07, Nicolas KOWALSKI <[EMAIL PROTECTED]> wrote:
> Hello,
>
> I noticed that "no_comp_alerts" does not work in the following configuration:
>
> watch HOSTGROUP
> service SVC
> interval 5m
> monitor MONITOR
> period failure: wd{mon-fri} hr{8-18}
> alert wh-file.alert
> period up: wd{mon-sun}
> no_comp_alerts
> upalert wh-file.alert
>
>
> Looking in the code and logs, I found that
> _no_comp_alerts_upalert_sent is never reset to zero when no alert
> script is called. So upalerts are sent only once (at startup or
> reset).
>
> The following patch against mon 1.2.0 seems to correct this, by
> forcing the reset as soon as possible:
>
> --- mon.orig 2007-07-23 17:37:10.988935200 +0200
> +++ mon 2007-07-23 17:38:34.916224800 +0200
> @@ -666,6 +666,18 @@
>
> my $pref = \%{$sref->{"periods"}->{$periodlabel}};
>
> + #
> + # Reset no_comp_alerts counter if ALERT
> + #
> + if ($pref->{"no_comp_alerts"} &&
> + $pref->{"_no_comp_alerts_upalert_sent"} > 0 &&
> + !($flags &
> ($FL_UPALERT|$FL_ACKALERT|$FL_DISABLEALERT|$FL_ENABLEALER
> +T))) {
> + syslog('debug',"$group/$service/$periodlabel: ALERT,
> no_comp_alerts
> +reset");
> + $pref->{"_no_comp_alerts_upalert_sent"} = 0;
> + }
> +
> #
> # skip upalerts/ackalerts not paired with down alerts
> # disable by setting "no_comp_alerts" in period section
>
>
> Does it look reasonable ?
Anyone about it ?
Without this patch, we have to use the following workaround, which is
a bit ugly:
watch HOSTGROUP
service SVC
interval 5m
monitor MONITOR
period failure: wd{mon-fri} hr{8-18}
alert wh-file.alert
period up: wd{mon-sun}
no_comp_alerts
alert fake.alert
upalert wh-file.alert
fake.alert is an alert script doing nothing.
Thanks,
--
Nicolas
_______________________________________________
mon mailing list
[email protected]
http://linux.kernel.org/mailman/listinfo/mon
