Hi All.

I have two running servers (drbd primary, drbd secondary). When there is a
problem with one of them we switch to the second one. Currently on the
/dev/drbd0 (it has /xxx ext3 filesystem directly on it) we have some
filesystem problems (after fsck -n). I would like to correct these errors
but the filesystem should be unmounted first (I would like to avoid it if
possible).

Is there a way to make fsck -a on mounted filesystem for ext3? My setup is:

ext3 filesystem (problems)
^
DRBD
^
LVS
^
RAID

Any ideas?

Best regards,
Rafal.
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos

On 04/12/2012 09:50 AM, Rafał Radecki wrote:
> Hi All.
>
> I have two running servers (drbd primary, drbd secondary). When there is a
> problem with one of them we switch to the second one. Currently on the
> /dev/drbd0 (it has /xxx ext3 filesystem directly on it) we have some
> filesystem problems (after fsck -n). I would like to correct these errors
> but the filesystem should be unmounted first (I would like to avoid it if
> possible).
>
> Is there a way to make fsck -a on mounted filesystem for ext3? My setup is:

No. None of the popular Filesystems support online fsck.

Regards,
Dennis
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos

--On Thursday, April 12, 2012 03:17:37 PM +0200 Dennis Jacobfeuerborn
<dennisml@conversis.de> wrote:

> On 04/12/2012 09:50 AM, Rafał Radecki wrote:
>>
>> I have two running servers (drbd primary, drbd secondary). When there is
>> a problem with one of them we switch to the second one. [...]
>>
>> Is there a way to make fsck -a on mounted filesystem for ext3?
>
> No. None of the popular Filesystems support online fsck.

However, if you're using pacemaker the RA (IIRC) has the option
of doing an 'fsck -p' before mounting. OTOH, since an fsck has
no predicatable run time limit, a long fsck might cause a
'failed start' condition, causing a cascade from one node to
another to another ...

(more details on this mechanism are probably more appropriate to the
applicable HA lists)

Devin

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos

On 04/12/2012 08:17 PM, Devin Reade wrote:
> --On Thursday, April 12, 2012 03:17:37 PM +0200 Dennis Jacobfeuerborn
> <dennisml@conversis.de> wrote:
>
>> On 04/12/2012 09:50 AM, Rafał Radecki wrote:
>>>
>>> I have two running servers (drbd primary, drbd secondary). When there is
>>> a problem with one of them we switch to the second one. [...]
>>>
>>> Is there a way to make fsck -a on mounted filesystem for ext3?
>>
>> No. None of the popular Filesystems support online fsck.
>
> However, if you're using pacemaker the RA (IIRC) has the option
> of doing an 'fsck -p' before mounting. OTOH, since an fsck has
> no predicatable run time limit, a long fsck might cause a
> 'failed start' condition, causing a cascade from one node to
> another to another ...
>
> (more details on this mechanism are probably more appropriate to the
> applicable HA lists)

"doing an 'fsck -p' before mounting" still means you have to take the fs
offline though so this doesn't really change anything.

Regards,
Dennis
_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos

--On Thursday, April 12, 2012 10:38:56 PM +0200 Dennis Jacobfeuerborn
<dennisml@conversis.de> wrote:

> On 04/12/2012 08:17 PM, Devin Reade wrote:

>> However, if you're using pacemaker the RA (IIRC) has the option
>> of doing an 'fsck -p' before mounting. [...]
>
> "doing an 'fsck -p' before mounting" still means you have to take the fs
> offline though so this doesn't really change anything.

Correct; it still has to be offline. I was mentioning
it as a possible alternative for the OP's use case if this is
more than just a "one of". On re-reading the original message,
though, it sounds like it might just be bare DRBD rather than
in combination with pacemaker.

Devin

_______________________________________________
CentOS mailing list
CentOS@centos.org
http://lists.centos.org/mailman/listinfo/centos