[Lustre-discuss] Force read-only ost mount

[Zachary Beebleson]

Is there a way to manually mount an ost read-only? I tried 'mount -t
lustre -o ro ost_device mountpoint', but got a

mount.lustre: mount /dev/data1recov/ost1 at /mnt/os1brecov failed:
Transport endpoint is not connected

I know if lustre detects a serious enough problem, ldiskfs error say, it
will remount the ost read-only, but I would like to manually force this.

A raid logical volume on an oss has some problems and I'm trying to
transfer one of two osts to another oss (I believe this is possible by
umounting lustre from all servers/clients, executing a 'tunefs.lustre
--writeconf' on all lustre servers and the related devices, and
remounting --- with the copied ost replacing the problematic one). In
order to minimize downtime, I would like to work with a static ost image
-- so I can transfer a faithful copy back, after I repair the raid
problem. I did deactivate the ost device on the mds, but I'm worried about
changes to files that live on the ost (I wasn't planning on creating
another image).

Thanks,
Zach
_______________________________________________
Lustre-discuss mailing list
Lustre-...@lists.lustre.org
http://lists.lustre.org/mailman/listinfo/lustre-discuss

[Andreas Dilger]

On 2011-06-29, at 10:37 AM, Zachary Beebleson wrote:
> Is there a way to manually mount an ost read-only? I tried 'mount -t
> lustre -o ro ost_device mountpoint', but got a
>
> mount.lustre: mount /dev/data1recov/ost1 at /mnt/os1brecov failed:
> Transport endpoint is not connected
>
> I know if lustre detects a serious enough problem, ldiskfs error say, it
> will remount the ost read-only, but I would like to manually force this.

There was a patch in bugzilla related to this, but it likely needs to be
updated. Some of the mount-time code tries to write to the filesystem,
and that needs to be disabled.

> A raid logical volume on an oss has some problems and I'm trying to
> transfer one of two osts to another oss (I believe this is possible by
> umounting lustre from all servers/clients, executing a 'tunefs.lustre
> --writeconf' on all lustre servers and the related devices, and
> remounting --- with the copied ost replacing the problematic one).
>
> In order to minimize downtime, I would like to work with a static ost image
> -- so I can transfer a faithful copy back, after I repair the raid
> problem. I did deactivate the ost device on the mds, but I'm worried about
> changes to files that live on the ost (I wasn't planning on creating
> another image).

Cheers, Andreas
--
Andreas Dilger
Principal Engineer
Whamcloud, Inc.