*** (0.28) *** Re: [Linux-HA] quorum server not available

[Andreas Kurz]

On 2/24/07, Alan Robertson <alanr at unix.sh> wrote:
> Andreas Kurz wrote:
> > On 2/23/07, Alan Robertson <alanr at unix.sh> wrote:
> >> Sebastian Reitenbach wrote:
> >> > Hi Alan,
> >> >
> >> >>> this is my ha.cf:
> >> >>> autojoin any
> >> >>> crm true
> >> >>> node ppsdb101
> >> >>> node ppsdb102
> >> >>> node ppsnfs101
> >> >>> node ppsnfs102
> >> >>> node ppsbackup101
> >> >>> respawn root /usr/lib64/heartbeat/quorumd
> >> >>> apiauth quorum uid=hacluster,root
> >> >>> mcast bridge0 224.0.0.0 694 2 0
> >> >>> cluster ppscluster
> >> >>> quorum_server quorum
> >> >>>
> >> >>>
> >> >> A comment:
> >> >>      If you have to run the quorumd as root, then something is
> >> >>      broken. DO NOT run things as root.  The only components in
> >> >>      heartbeat which should need to run as root are heartbeat itself
> >> >>      and the lrmd.  Please do not run anything else as root.
> >> > I tested with the following lines in the ha.cf file:
> >> > respawn hacluster /usr/lib64/heartbeat/quorumd
> >> > apiauth quorum uid=hacluster
> >> >
> >> > but then I saw the logs, that the quorum was started, and then some
> >> short time
> >> > later died
> >> > with an error message, see attached log file.
> >> >
> >> >> Another note:
> >> >>      The quorumd will never do anything helpful in a 5-node
> >> >>      cluster.  What do you think it will do for you here?
> >> > I tried this, removed the three quorum lines in the ha.cf file on
> >> all nodes, and
> >> > restarted
> >> > heartbeat on them, then all five nodes came up (green dots in the
> >> gui), but in
> >> > the hb_gui,
> >> > the cluster was without quorum.
> >>
> >> Feb 23 19:12:53 ppsnfs101 quorumd: [7565]: ERROR: already running:
> >> [pid -3].
> >> Feb 23 19:12:53 ppsnfs101 quorumd: [7565]: ERROR: Startup aborted
> >> (already running).Shutting down.
> >>
> >>
> >> Well... This claims that i was already running - but I have no idea why
> >> it gave a pid of -3
> >
> > I am also doing some tests with a two-node cluster and quorumd on a
> > third node. What i found out:
> >
> > * quorumd has to be started as root, because it insists on creating a
> > pid file in /var/run .... normally the user haclient has not enough
> > rights to do this .... the error message is misleading
>
> Sounds like we ought to fix this!  Do you want to make a bugzilla for
> this for us?

I've added a bug to bugzilla.

>
> > * if I understood right the main idea of quorumd is to run it _NOT_ on
> > a cluster member ... I start it out of the inittab on an extra server,
> > let init respawn it in case of an error ... works OK
> >
> > * when the cluster is up, only the DC communicates with the external
> > quorumd
>
> Exactly!
>
>
> --
>     Alan Robertson <alanr at unix.sh>
>
> "Openness is the foundation and preservative of friendship...  Let me
> claim from you at all times your undisguised opinions." - William
> Wilberforce
> _______________________________________________
> Linux-HA mailing list
> Linux-HA at lists.linux-ha.org
> http://lists.linux-ha.org/mailman/listinfo/linux-ha
> See also: http://linux-ha.org/ReportingProblems
>