[Pacemaker] ocfs2_controld: Unable to connect to CKPT: Object does not exist

[Daniel Qian]

----- Original Message ----- 
From: "Jiaju Zhang" <jjzhang.linux at gmail.com>
To: "Daniel Qian" <daniel at bestningning.com>
Sent: Thursday, December 31, 2009 3:27 AM
Subject: Re: [Pacemaker] ocfs2_controld: Unable to connect to CKPT: Object 
does not exist


> On Thu, Dec 31, 2009 at 3:57 PM, Daniel Qian <daniel at bestningning.com> 
> wrote:
>>
>> ----- Original Message ----- From: "Jiaju Zhang" 
>> <jjzhang.linux at gmail.com>
>> To: <pacemaker at oss.clusterlabs.org>; "Daniel Qian" 
>> <daniel at bestningning.com>
>> Sent: Thursday, December 31, 2009 2:26 AM
>> Subject: Re: [Pacemaker] ocfs2_controld: Unable to connect to CKPT: 
>> Object
>> does not exist
>>
>>
>> On Thu, Dec 31, 2009 at 11:35 AM, Daniel Qian <daniel at bestningning.com>
>> wrote:
>>>
>>> I am using pacemaker, corosync and ocfs2 on Fedora 12 to build an
>>> active/active cluster. When I try to start up o2cb resource with
>>> ocfs2-tools-pcmk-1.4.3-3.fc12.x86_64 that comes with Fedora 12 it 
>>> produces
>>> the following errors:
>>>
>>> Dec 30 22:06:29 ilo150 corosync[3866]: [pcmk ] info: pcmk_notify:
>>> Enabling node notifications for child 9178 (0x1e5e5f0)
>>> Dec 30 22:06:29 ilo150 ocfs2_controld[9178]: Unable to connect to CKPT:
>>> Object does not exist
>>>
>>> Does anyone know if this is a problem on the ocfs2 side or corosync 
>>> side?
>>
>> - Have you installed openais package?
>> - ocfs2_controld is using the CKPT service which is provided by openais
>> - not corosync.
>>
>> really? I did not know that. So my best bet is either switch to openais 
>> or
>
> You might misunderstand what I said. It is not "switch to openais" but 
> "install
> the openais as well". That is to say, both corosync and openais package
> are needed :)
>

I am getting a little confused now and maybe I missed some information here. 
On the other cluster where openais is in use I had to set the cluster 
communication - authkey, totem and interface etc. With this one I am working 
on I did pretty much the same in corosync. Now you are saying I need to turn 
on openais to work with corosync but the question is where I should put 
those settings? openais.conf, corosync.conf or both? and which package is 
supposed to start up pacemaker processes which are done through corosync at 
the moment:

 1111 ?        Ssl    0:10 corosync
 1117 ?        SLs    0:00  \_ /usr/lib64/heartbeat/stonithd
 1118 ?        S      0:03  \_ /usr/lib64/heartbeat/cib
 1119 ?        S      0:01  \_ /usr/lib64/heartbeat/lrmd
 1120 ?        S      0:00  \_ /usr/lib64/heartbeat/attrd
 1121 ?        S      0:00  \_ /usr/lib64/heartbeat/pengine
 1122 ?        S      0:02  \_ /usr/lib64/heartbeat/crmd



>> to use gfs with corosync then. I will try it out and post the result 
>> here.
>
> But gfs2 uses dlm as the lock manager, and dlm also uses CKPT service. So
> eventually you also need the openais package.
>


DLM seems to have started fine in my cluster

[root at ilo143 ~]# crm_mon -1


============
Last updated: Thu Dec 31 11:12:35 2009
Stack: openais
Current DC: ilo143 - partition with quorum
Version: 1.0.5-ee19d8e83c2a5d45988f1cee36d334a631d84fc7
2 Nodes configured, 2 expected votes
3 Resources configured.
============

Online: [ ilo143 ilo150 ]

 Master/Slave Set: drbd_clone0
     Masters: [ ilo143 ]
     Slaves: [ ilo150 ]
 Clone Set: dlm-clone
     Started: [ ilo143 ilo150 ]

Failed actions:
    o2cb:1_start_0 (node=ilo150, call=22, rc=1, status=complete): unknown 
error
    o2cb:0_start_0 (node=ilo143, call=21, rc=1, status=complete): unknown 
error

Basically I have been following the doc "Cluster from Scratch Apache in 
Fedora 11" at clusterlabs.org website only I am on Fedora 12


Thanks,
Daniel