[Linux-cluster] OCFS2 and SAN MultiPath I/O

[Flavio Junior]

Ok.. seems like I found my real problem..

I'm using OCFS2 Threshold as 30 seconds (value 61) and my multipath
RDAC devloss timeout was 60 seconds.

Adjust these values seems to solve the problem with incorrect fencing.


Any news I report here, thanks :)

--

Flávio do Carmo Júnior aka waKKu

On Wed, Apr 22, 2009 at 4:10 PM, Flavio Junior <billpp at gmail.com> wrote:
> On Wed, Apr 22, 2009 at 2:50 PM, Joel Becker <Joel.Becker at oracle.com> wrote:
>>
>>        What's the oops look like?  You shouldn't get an oops from a
>> path going down.  If you do, of course the machine is going to reboot -
>> it's in a broken state.  You're using RDAC, not dm-multipath, so I'm
>> betting you have to ask IBM why their multipath code is oopsing.
>>        As a short sanity check, make sure ocfs2 is mounted on the
>> multipath device, not a component path.
>>
>
> Hi Joel, thanks for the answer...
>
> OK, but i'm not sure about a kernel oops being executed.
>
> I'd change panic_on_oops to 0 and it work "as expected" (well, you
> will see that I wasnt totally sincere about my setup. I'm using RHCS
> and gfs too :X)
>
> Here is the paste of /var/log/messages from the remaining node
> (panic_on_oops=0) when I turn off the fiber switch for primary path.
>
> http://rafb.net/p/TgLoWH71.html
>
>
>> Joel
>>
>
>
> Thanks again,
>
> --
>
> Flávio do Carmo Júnior aka waKKu
>