[Linux-HA] OCFS2 on HB 2.1.3 v2

[Christian Rishøj]

On 18 Feb 2008, at 05:45, Michael Brennen wrote:

> On Mon, 18 Feb 2008, Christian Rishøj wrote:
>
>> I had similar issues with OCFS2 on DRBD: Undeletable files/folders,  
>> inability to reboot. Furthermore, I saw some quite severe  
>> performance issues with OCFS2 in comparison with XFS or EXT3.
>
> Hm... good feedback.  Were you also running OCFS2 in dual master  
> mode, or master/slave?

Dual master. However, as I recall it, the performance issues were  
present even when the other node was down.


> ext3 is not a cluster file system such that a single network mount  
> can be safely accessible from multiple computers at once (cache  
> coherency, locking.)  I've not used xfs, but I did not recall that  
> it is a cluster file system.  Has that changed?

In our current setup, the filesystem is only used by a single active  
node at a time.



>> On Feb 18, 2008 1:29 AM, Eddie C <edlinuxguru at gmail.com> wrote:
>>> I tried this configuration. I found that the active/active
>>> configuration worked but had strange issues. These may have been
>>> related to the stability of OCFS2 itself or it might have had to do
>>> with running on a non supported OS (Fedora Core 5). I had strange
>>> issues. Folders that would not delete, system would not reboot after
>>> OCFS2 hang. I spoke with someone else who had similar issues getting
>>> this type of system into production.
>>>
>>> I would suggest having with the active/passive replicated disk
>>> formatted with ext3 as a backup plan.
>>>
>>> On Feb 17, 2008 2:23 PM, Michael Brennen <michael at fishnet.us> wrote:
>>>>
>>>> Can someone give pointers to integrating ocfs2 with heartbeat? The
>>>> idea is to run ocfs2 as the cluster file system on the real servers
>>>> running on an iscsi failover backend cluster.  Apparently some
>>>> userspace patches are required to ocfs2 to let hb manage it, but I
>>>> think things have changed much since most of the information I've
>>>> found with google was published.
>>>>
>>>> ocfs2 seems to be the best available cluster file system to run  
>>>> with
>>>> hb.  The gfs file system is perhaps the most complete, but it is  
>>>> all
>>>> entangled with the redhat clvm and cluster management, and I do not
>>>> want to manage that.  Does anyone have a better suggestion than
>>>> ocfs2?
>>>>
>>>>    -- Michael
>>>> _______________________________________________
>>>> 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
>>
>
>   -- Michael_______________________________________________
> 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