Hello,

On 02/22/2012 11:37 PM, Tom Hanstra wrote:
> See my further information with TH>> below...
> 
> On 02/22/2012 05:09 PM, Andreas Kurz wrote:
>> Hello Tom,
>>
>> On 02/22/2012 08:01 PM, Tom Hanstra wrote:
>>> I'm new to the Linux-HA clustering, though I've had experience
>>> with RedHat's Cluster packages for several years. I'm trying to see how
>>> the open source software compares.
>>>
>>> So, I set up two KVM Virtual servers running RHEL6 and compiled and
>>> installed the Cluster Glue, Heartbeat, and Pacemaker software. I was
>>> able to get two nodes running, though there are some errors which I will
>>> need to track down.
>> Oh ... why did you build the complete stack manually? Pacemaker is
>> technology preview in RHEL6 and it ships latest version .... in
>> combination with corosync instead of Heartbeat this works really fine.
>>
> TH>> Unfortunately, I'm limited to the educational version of RHEL6 
> which does not include any of the clustering software without additional 
> charges.  I just did a check on both corosync and pacemaker.  For 
> corosync, the packages show up but are inaccessible; for pacemaker, only 
> pacemaker-cts is available.  I'm not sure if this is sufficient but 
> doubt it.

I see ... well, Centos and Scientific Linux have all packages in their
repos ...

> 
> But is corosync better than heartbeat?  Or am I getting into a religious 
> war by asking that?

Since heartbeat is not actively developed any more, corosync is the way
to go for a future proof setup.

> 
> 
>>>    From my other cluster experience, I know that getting fencing/stonith
>>> set up properly is something necessary and I want to work on that even
>>> before I try to track down other problems further. Without the ability
>>> to kill off a node, odd things can happen. So, my focus right now is on
>>> finding a working stonith device for this setup.
>>>
>>> I got all of the pieces I think I need for the external/libvirt device,
>>> have fence_virtd running on the host box and I do get output on both
>>> host and clients from the fence_xvm command:
>>>
>>> 1023$ fence_xvm -o list
>>> RH5_LIS0 25132742-8e3a-a1f2-a862-de3705ea8d8f on
>>> RH5_LIS1 2b4d4813-0107-6aec-a66f-2159ec95da4c on
>>> RH5_LIS2 fa6e2603-f7d6-34fa-dd03-4886cdf6e44b on
>>> RH5_LIS3 aafc6639-2d29-8bbe-4d62-38498f390563 on
>>> RH6_WITS7 51c15635-889f-7213-b0e2-e213f771e52a on
>>> RH6_WITS8 2ffdfebe-d49b-698b-b76c-a4abd8cbf42a on
>>>
>>> Where I am running into problems right now is translating the
>>> information I have from this command into the proper setup and syntax to
>>> set this as a stonith device and actually test killing off a node. The
>>> information given by this command gives the names of the virtual
>>> machines. But in my cluster setup, I have given these node names:
>>>
>>> lv7-eli = RH6_WITS7
>>> lv8-eli = RH6_WITS8
>> Try something like that for a single host setup:
>>
>> primitive stonith_lv7-eli stonith:fence_virt \
>>      params pcmk_host_check="static-list" \
>>      pcmk_host_list="lv7-eli" \
>>      port="RH6_WITS7" \
>>      op monitor interval="600s"
>>
> TH>> Bear with me a bit.  This is a crm configuration command, right.  
> Can you help me understand where the information gets stored when I 
> issue this command?  I was thinking it would go to a file somewhere, but 
> as you mention later, this information does not come from the ha.cf 
> file.  Where does it go?

The cib.xml file is stored in "/var/lib/heartbeat/crm" directory and
propagated to all nodes ... dont't manipulate it manually .... "crm
configure show" gives you the crm syntax version which is much easier to
read.

>> ... and the same for the other node with adopted names. You should also
>> take care to run the stonith resources not on that node that can be
>> fenced by it ... like:
>>
>> location l_stonith_lv7-eli stonith_lv7-eli -inf: lv7-eli
>>
> TH>> I'm not clear on what you mean here.  This another configuration 
> command, but I don't understand what it is doing.  In my two node 
> cluster, each node should be able to fence off the other.  How does this 
> command help to accomplish that?

This is a location constraints that disallows the stonith resource
capable of fencing lv7-eli to run on node lv7-eli.

> 
>>> What is the proper stonith command that will actually kill off a node in
>>> such a KVM setup? And how does that translate into settings I would add
>>> to my ha.cf file?
>> Even if you continue to use Heartbeat ccm instead of corosync, there is
>> nothing to be added to ha.cf, all stonith resource configuration is done
>> in the cib.
>>
>> Regards,
>> Andreas
>>
> Thanks for your help,

You are welcome!

> Tom
> 

Regards,
Andreas

-- 
Need help with Pacemaker?
http://www.hastexo.com/services/custom-training


Attachment: signature.asc
Description: OpenPGP digital signature

_______________________________________________
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