On Fri, Jun 20, 2008 at 01:33:40PM +0200, Andrew Beekhof wrote:
> On Fri, Jun 20, 2008 at 12:35, Michael Alger <[EMAIL PROTECTED]> wrote:
>> Is it possible to configure heartbeat such that a managed
>> resource (an IPaddr) is affected by the status of an unmanaged
>> resource?
> 
> yes - this was the original purpose of "unmanaged".
> although i think that we don't start monitors for unmanaged
> resources (which if true, would be a bug)

Well, it definitely appears to be true. It runs the monitor command
on both hosts (in my 2 node cluster, so presumably that means "all
hosts") to get the current status when it first starts up, and
crm_mon will report something like:

ipaddr_routing  (heartbeat::ocf:IPaddr):   Started  proxy-02
squid   (local::ocf:squid)[     proxy-02  proxy-01 ]

However it never seems to execute the monitor script again: if I
shut down squid on one server it remains listed as running on both,
even after a few days (I left it shutdown over the weekend). This
means that even though I have a constrain to keep squid with the
IPaddr, when I caused pingd to fail on the host where squid is
running, the address was moved to the host were squid was stopped.
Heartbeat never ran the monitor command again.

Another issue is that heartbeat is unable to shut down if you have
an unmanaged resource started when heartbeat starts. It logs this
message once every 60 seconds:

crmd[2465]: 2008/06/23_12:52:28 info: do_shutdown: All subsystems
stopped, continuing
crmd[2465]: 2008/06/23_12:52:28 ERROR: verify_stopped: Resource
squid was active at shutdown.  You may ignore this error if it is
unmanaged.

Even manually stopping squid doesn't let it shutdown, as it never
actually checks the status. Killing the crmd process resulted in the
shutdown of heartbeat continuing.

I've added an entry for this to bugzilla, bug 1926:

http://developerbugs.linux-foundation.org/show_bug.cgi?id=1926
_______________________________________________
Linux-HA mailing list
[email protected]
http://lists.linux-ha.org/mailman/listinfo/linux-ha
See also: http://linux-ha.org/ReportingProblems

Reply via email to