[Linux-HA] Two node, two IP resource, config problem?

[Robert Gravsjö]

Andrew Beekhof wrote:
> On 9/8/06, Robert Gravsjö <robert.gravsjo at tietoenator.com> wrote:
>>
>>
>> Oren Nechushtan wrote:
>> > Did you try the patch Andrew supplied? It worked for me.
>> > See
>> http://lists.community.tummy.com/pipermail/linux-ha/2006-August/021391.html
>>
>> >
>> > 13 days agoFilter out updates that arent for cluster members (eg.
>> ping nodes)
>> >
>> > changeset
>> > Andrew Beekhof <beekhof at gmail.com> [Thu, 24 Aug 2006 17:42:40 +0200]
>> rev 9564
>> >
>> > Filter out updates that arent for cluster members (eg. ping nodes)
>>
>> I tried this patch and it fixed the failover problem.
>>
>> The OFFLINE problem still occurs. A node with its network cables removed
>> will stay OFFLINE in crm_mon despite the fact that it is back online.
> 
> How can the other node(s) know the "failed" node is "back" if its
> network cables are unplugged?

There are additional networks attached to these servers. One of these
other networks are dedicated to heartbeat communication.
So, since heartbeat is able to communicate with the other node it should
also be aware of it resuming normal operations. The log shows entries
about it being back, but crm_mon shows it OFFLINE and crmadmin says "idle".

> 
>> The difference this time is that restarting heartbeat results in both
>> node starting up group net_1 and no one runs group_1.
> 
> Is this a two node cluster?

Yes, this is a two node cluster sharing two IP addresses (one for the
primary network and one for the secondary).

/roppert

> 
>>
>> Is any v2.x ready for production environment? Or should I revert back to
>> v1 if I want to use heartbeat in production environment?
>>
>>
>> Regards,
>> roppert
>>
>>
>> >
>> > Oren.
>> >
>> >> -----Original Message-----
>> >> From: linux-ha-bounces at lists.linux-ha.org
>> >> [mailto:linux-ha-bounces at lists.linux-ha.org]On Behalf Of
>> >> robert.gravsjo at tietoenator.com
>> >> Sent: Thursday, September 07, 2006 2:03 PM
>> >> To: General Linux-HA mailing list
>> >> Subject: [Linux-HA] Two node, two IP resource, config problem?
>> >>
>> >>
>> >> Hi,
>> >>
>> >> I having trouble understanding the behavior I see when using the
>> >> attached configuration.
>> >> The scenario is pretty simple: as long as net_1 has some connectivity
>> >> stay on active node else failover net_1 and group_1 to the other node.
>> >>
>> >> Now, the strangeness is that if I use ifconfig to bring my interfaces
>> >> down the failover occurs as expected but if I physicly pull
>> >> the plug on
>> >> the active nodes interfaces the failover does not occur. Insted the
>> >> active node is set to OFFLINE and all resources in group_1 is
>> >> stopped on
>> >> both nodes. I can't understand why?
>> >>
>> >> The second thing is that after a failover heartbeat doesnt accept the
>> >> failed node even if it comes back online. crm_mon still shows
>> >> that node
>> >> as OFFLINE. Restarting heartbeat manually on that node makes it online
>> >> again. I'm not sure why this happens either?
>> >>
>> >> Best regards,
>> >> roppert
>> >>
>> >> --
>> >> RobertG
>> >>
>> > _______________________________________________
>> > 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
>> >
>>
>> -- 
>> RobertG
>>
>> Phone: +46 (0)480 44 58 35
>> _______________________________________________
>> 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
>>
> _______________________________________________
> 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
> 

-- 
RobertG

Phone: +46 (0)480 44 58 35