[Pacemaker] dampen / pingd - how to be sure pingd will be updated

Andrew Beekhof andrew at beekhof.net
Mon Jan 17 09:49:52 UTC 2011


On Tue, Dec 21, 2010 at 7:52 AM, Thomas Guthmann <tguthmann at iseek.com.au> wrote:
> Hey guys,
>
> This may be dumb or obvious but it took me a long time to understand why my pingd with
> dampen never got updated! So I think that this may be useful for everybody to share.
> In short:
>
>  " You MUST define a monitor interval HIGHER than the the dampen delay. "

Which version was this?
I seem to recall fixing something similar to this (its a bug in attrd)
but 1.0.10 should be ok.

>
> An example is better than a long speech :
>
> *WILL WORK*
> primitive pingd ocf:pacemaker:ping \
>        params host_list="10.1.1.1" attempts="5" timeout="2" debug="true" dampen="50s" \
>        op monitor interval="60" timeout="60" \
>        op start interval="0" timeout="90" \
>        op stop interval="0" timeout="90"
>
> Because Monitor interval > dampen (60 > 50)
>
> *WILL NOT WORK*
> primitive pingd ocf:pacemaker:ping \
>        params host_list="10.1.1.1" attempts="5" timeout="2" debug="true" dampen="50s" \
>        op monitor interval="10" timeout="60" \
>        op start interval="0" timeout="90" \
>        op stop interval="0" timeout="90"
>
> Because Monitor interval < dampen (10 < 50)
>
> *Explanations*
> Why ? Because every 10sec (monitor interval) pacemaker will trigger attrd_updater
> (check your logfiles for [1]). attrd_updater will wait for the dampen time given as
> an argument (-d 50s in our example). However we won't be able to wait so long because
> 10sec later, attrd_updater is called again and obviously it resets dampen again and
> again. Then pingd will never reach its dampen value and consequently pacemaker will
> NEVER update pingd (except if you force a reset with attrd_updater -R or if you
> modify the CIB). Q.E.D.
>
> Something somewhere should test if dampen is inferior than monitor time. It could be
> checked in the OCF in ping_validate(). But it doesn't seem to be used (note that their is
> code for an interval parameter that doesn't exist in ocf:pacemaker:ping anyway but comes
> from the old ocf:pacemaker:pingd).
>
> Any objections/comments to this deductive reasoning ?
>
> [1] : attrd_updater: [9712]: info: Invoked: attrd_updater -n pingd -v 0 -d 50s
>
> -Thomas
>
> _______________________________________________
> Pacemaker mailing list: Pacemaker at oss.clusterlabs.org
> http://oss.clusterlabs.org/mailman/listinfo/pacemaker
>
> Project Home: http://www.clusterlabs.org
> Getting started: http://www.clusterlabs.org/doc/Cluster_from_Scratch.pdf
> Bugs: http://developerbugs.linux-foundation.org/enter_bug.cgi?product=Pacemaker
>
>




More information about the Pacemaker mailing list