[Cluster-devel] pcmk+corosync controld,dlm support without cman

[Cluster-devel] pcmk+corosync controld,dlm support without cman

[Nick Khamis]

That would be really great. Is there any talk of when the next stable
release of:
* pcmk+corosync supporting direct communication with dlm gfs, controld
* The separation of dlm, gfs, controld from cman

will be availabile?

Thanks in Advance,

Nick.

On Mon, Oct 31, 2011 at 4:52 AM, Andrew Beekhof <andrew@beekhof.net> wrote:
> On Sat, Oct 29, 2011 at 3:09 AM, Nick Khamis <symack@gmail.com> wrote:
>> Hello Gents,
>>
>> Thank you so much for your response. That being said, what are the
>> plans once the next
>> release of CMAN does not include PCMK and DLM related implementation? From what
>> I can see, libdlm will be separated from the project completely.
>
> When CMAN goes away, the dlm gfs controld's will be talking directly
> to corosync.
> I believe the work has already been done but may need additional testing.
>
>>
>> Thanks in Advance,
>>
>> Nick.
>>
>> On Fri, Oct 28, 2011 at 3:34 AM, Vladislav Bogdanov
>> <bubble@hoster-ok.com> wrote:
>>> 28.10.2011 04:04, Nick Khamis wrote:
>>>> Hello Everyone,
>>>>
>>>> I just want to make sure this is still the case before I go through
>>>> with it. I am trying to setup an
>>>> active/active using:
>>>>
>>>> Corosync 1.4.2
>>>> Pacemaker 1.1.6
>>>> Cluster3
>>>> DRBD 8.3.7
>>>> OCFS2
>>>>
>>>> The only reason I installed Cluster3 was for dlm support. Fencing
>>>> would be handled by stonithd.
>>>> Is there any reason I need to fully install and configure two cluster
>>>> managers. Defining nodes in
>>>> both cman and corosync/openais?
>>>
>>> You may look at
>>> http://www.mail-archive.com/pacemaker at oss.clusterlabs.org/msg09959.html
>>>
>>> That message contains Andrew's dlm_controld patch for shortening fencing
>>> path - it now talks directly to stonith subsystem, rebased for 3.0.17,
>>> and some fixes to make it all compile. Some work should probably be done
>>> to make it apply/compile on latest cluster3.
>>>
>>> That way I eliminate cman from stack (except some dlm-related libs) and
>>> run cluster with stable fencing on top of corosync/openais.
>>>
>>> Best,
>>> Vladislav
>>> _______________________________________________
>>> 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
>>
> _______________________________________________
> 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
>

[Cluster-devel] pcmk+corosync controld, dlm support without cman

[Steven Dake]

On 11/01/2011 06:23 PM, Nick Khamis wrote:
> That would be really great. Is there any talk of when the next stable
> release of:
> * pcmk+corosync supporting direct communication with dlm gfs, controld
> * The separation of dlm, gfs, controld from cman
> 
> will be availabile?
> 
> Thanks in Advance,
> 
> Nick.
> 
> On Mon, Oct 31, 2011 at 4:52 AM, Andrew Beekhof <andrew@beekhof.net> wrote:
>> On Sat, Oct 29, 2011 at 3:09 AM, Nick Khamis <symack@gmail.com> wrote:
>>> Hello Gents,
>>>
>>> Thank you so much for your response. That being said, what are the
>>> plans once the next
>>> release of CMAN does not include PCMK and DLM related implementation? From what
>>> I can see, libdlm will be separated from the project completely.
>>
>> When CMAN goes away, the dlm gfs controld's will be talking directly
>> to corosync.
>> I believe the work has already been done but may need additional testing.
>>
>>>
>>> Thanks in Advance,
>>>
>>> Nick.
>>>
>>> On Fri, Oct 28, 2011 at 3:34 AM, Vladislav Bogdanov
>>> <bubble@hoster-ok.com> wrote:
>>>> 28.10.2011 04:04, Nick Khamis wrote:
>>>>> Hello Everyone,
>>>>>
>>>>> I just want to make sure this is still the case before I go through
>>>>> with it. I am trying to setup an
>>>>> active/active using:
>>>>>
>>>>> Corosync 1.4.2
>>>>> Pacemaker 1.1.6
>>>>> Cluster3
>>>>> DRBD 8.3.7
>>>>> OCFS2
>>>>>
>>>>> The only reason I installed Cluster3 was for dlm support. Fencing
>>>>> would be handled by stonithd.
>>>>> Is there any reason I need to fully install and configure two cluster
>>>>> managers. Defining nodes in
>>>>> both cman and corosync/openais?
>>>>

The developer community is targeting Fedora 17 for upstream integration
of this work.  Fedora 17 schedule hasn't been published and won't be
until F16 is finalized.  (next few weeks).  I expect other distros will
follow on once the dust settles.

Regards
-steve
.
>>>> You may look at
>>>> http://www.mail-archive.com/pacemaker at oss.clusterlabs.org/msg09959.html
>>>>
>>>> That message contains Andrew's dlm_controld patch for shortening fencing
>>>> path - it now talks directly to stonith subsystem, rebased for 3.0.17,
>>>> and some fixes to make it all compile. Some work should probably be done
>>>> to make it apply/compile on latest cluster3.
>>>>
>>>> That way I eliminate cman from stack (except some dlm-related libs) and
>>>> run cluster with stable fencing on top of corosync/openais.
>>>>
>>>> Best,
>>>> Vladislav
>>>> _______________________________________________
>>>> 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
>>>
>> _______________________________________________
>> 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
>>
> 

[Cluster-devel] pcmk+corosync controld, dlm support without cman

[Nick Khamis]

Hello Steve,

Thank you so much for your response. With the help of LinuxHA and ClusterDev
we have been lucky with building the stable versions of Cluster 3.1.7,
PCMK 1.1.6,
Corosync/OpenAIS 1.4.2. Almost offering active/active using the
mentioned stack on
ocfs2. Currently we are experiencing some problems with
ocfs2_controld.pcmk giving
the following error:

ocfs2_controld[14893]: 2011/11/02_11:38:55 info: crm_new_peer: Node 1
is now known as astdrbd1
ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: crm_abort:
send_ais_text: Triggered assert at ais.c:346 : dest != crm_msg_ais
Sending message 0 via cpg: FAILED (rc=22): Message error: Success (0)
ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: send_ais_text:
Sending message 0 via cpg: FAILED (rc=22): Message error: Success (0)
ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: crm_abort:
send_ais_text: Triggered assert at ais.c:346 : dest != crm_msg_ais
Sending message 1 via cpg: FAILED (rc=22): Message error: Success (0)
ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: send_ais_text:
Sending message 1 via cpg: FAILED (rc=22): Message error: Success (0)
1320248335 setup_stack at 170: Cluster connection established.  Local node id: 1
1320248335 setup_stack at 174: Added Pacemaker as client 1 with fd -1
root at astdrbd1:/usr/local/src/ocfs2-tools-1.6.4# /etc/init.d/o2cb status

Another concern is the redundancy found in using a cman+pcmk stack
(i.e., fenced vs. stonith, cman vs. pacemaker etc..).

Cheers,

Nick.




On Wed, Nov 2, 2011 at 3:29 PM, Steven Dake <sdake@redhat.com> wrote:
> On 11/01/2011 06:23 PM, Nick Khamis wrote:
>> That would be really great. Is there any talk of when the next stable
>> release of:
>> * pcmk+corosync supporting direct communication with dlm gfs, controld
>> * The separation of dlm, gfs, controld from cman
>>
>> will be availabile?
>>
>> Thanks in Advance,
>>
>> Nick.
>>
>> On Mon, Oct 31, 2011 at 4:52 AM, Andrew Beekhof <andrew@beekhof.net> wrote:
>>> On Sat, Oct 29, 2011 at 3:09 AM, Nick Khamis <symack@gmail.com> wrote:
>>>> Hello Gents,
>>>>
>>>> Thank you so much for your response. That being said, what are the
>>>> plans once the next
>>>> release of CMAN does not include PCMK and DLM related implementation? From what
>>>> I can see, libdlm will be separated from the project completely.
>>>
>>> When CMAN goes away, the dlm gfs controld's will be talking directly
>>> to corosync.
>>> I believe the work has already been done but may need additional testing.
>>>
>>>>
>>>> Thanks in Advance,
>>>>
>>>> Nick.
>>>>
>>>> On Fri, Oct 28, 2011 at 3:34 AM, Vladislav Bogdanov
>>>> <bubble@hoster-ok.com> wrote:
>>>>> 28.10.2011 04:04, Nick Khamis wrote:
>>>>>> Hello Everyone,
>>>>>>
>>>>>> I just want to make sure this is still the case before I go through
>>>>>> with it. I am trying to setup an
>>>>>> active/active using:
>>>>>>
>>>>>> Corosync 1.4.2
>>>>>> Pacemaker 1.1.6
>>>>>> Cluster3
>>>>>> DRBD 8.3.7
>>>>>> OCFS2
>>>>>>
>>>>>> The only reason I installed Cluster3 was for dlm support. Fencing
>>>>>> would be handled by stonithd.
>>>>>> Is there any reason I need to fully install and configure two cluster
>>>>>> managers. Defining nodes in
>>>>>> both cman and corosync/openais?
>>>>>
>
> The developer community is targeting Fedora 17 for upstream integration
> of this work. ?Fedora 17 schedule hasn't been published and won't be
> until F16 is finalized. ?(next few weeks). ?I expect other distros will
> follow on once the dust settles.
>
> Regards
> -steve
> .
>>>>> You may look at
>>>>> http://www.mail-archive.com/pacemaker at oss.clusterlabs.org/msg09959.html
>>>>>
>>>>> That message contains Andrew's dlm_controld patch for shortening fencing
>>>>> path - it now talks directly to stonith subsystem, rebased for 3.0.17,
>>>>> and some fixes to make it all compile. Some work should probably be done
>>>>> to make it apply/compile on latest cluster3.
>>>>>
>>>>> That way I eliminate cman from stack (except some dlm-related libs) and
>>>>> run cluster with stable fencing on top of corosync/openais.
>>>>>
>>>>> Best,
>>>>> Vladislav
>>>>> _______________________________________________
>>>>> 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
>>>>
>>> _______________________________________________
>>> 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
>>>
>>
>
>

[Cluster-devel] pcmk+corosync controld, dlm support without cman

[Andrew Beekhof]

On Thu, Nov 3, 2011 at 6:41 AM, Nick Khamis <symack@gmail.com> wrote:
> Hello Steve,
>
> Thank you so much for your response. With the help of LinuxHA and ClusterDev
> we have been lucky with building the stable versions of Cluster 3.1.7,
> PCMK 1.1.6,
> Corosync/OpenAIS 1.4.2. Almost offering active/active using the
> mentioned stack on
> ocfs2. Currently we are experiencing some problems with
> ocfs2_controld.pcmk giving
> the following error:
>
> ocfs2_controld[14893]: 2011/11/02_11:38:55 info: crm_new_peer: Node 1
> is now known as astdrbd1
> ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: crm_abort:
> send_ais_text: Triggered assert at ais.c:346 : dest != crm_msg_ais
> Sending message 0 via cpg: FAILED (rc=22): Message error: Success (0)
> ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: send_ais_text:
> Sending message 0 via cpg: FAILED (rc=22): Message error: Success (0)
> ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: crm_abort:
> send_ais_text: Triggered assert at ais.c:346 : dest != crm_msg_ais
> Sending message 1 via cpg: FAILED (rc=22): Message error: Success (0)
> ocfs2_controld[14893]: 2011/11/02_11:38:55 ERROR: send_ais_text:
> Sending message 1 via cpg: FAILED (rc=22): Message error: Success (0)
> 1320248335 setup_stack at 170: Cluster connection established. ?Local node id: 1
> 1320248335 setup_stack at 174: Added Pacemaker as client 1 with fd -1
> root at astdrbd1:/usr/local/src/ocfs2-tools-1.6.4# /etc/init.d/o2cb status
>
> Another concern is the redundancy found in using a cman+pcmk stack
> (i.e., fenced vs. stonith, cman vs. pacemaker etc..).

cman and pacemaker is an apples/oranges comparison, they fill
completely different roles as explained in Clusters from Scratch.
fenced and stonith is a duplication but one thats hard to avoid since
the "standard" dlm was written for fenced and pacemaker needs stonith
for other scenarios.

[Cluster-devel] pcmk+corosync controld, dlm support without cman

[Nick Khamis]

Yeah I did see that... It's all going to come together when cman is
retired, and the rest of the cluster master components are separated.


Cheers,

Nick.