Re: [Savannah-hackers-public] Re: [gnu.org #649274] colonialone ailing

[Sylvain Beucler]

On Mon, Dec 20, 2010 at 06:48:16PM -0500, Bernardo Innocenti via RT wrote:
> 
> REPLIES GO TO REQUESTORS BY DEFAULT.
> 
> <URL: http://rt.gnu.org/Ticket/Display.html?id=649274 >
> 
> On Mon, 2010-12-20 at 22:49 +0100, Sylvain Beucler wrote:
> 
> > I don't have a recommendation for the hardware issues.
> 
> To minimize the chance of data loss in case the problem persists, we
> would prefer to take the VMs down during resync as a precaution, 
> 
> What are your preferred times for 3-4 hours of planned maintenance
> downtime?

I won't be available much during the next few days, do I don't
personaly have a preferred time.

Only, make sure you temporarily disable the backup cron at
savannah-backup.gnu.org:/etc/cron.d/savannah-backup (and enable it
back afterwards).


> > - ward recommended using smaller RAID partitions used as separate LVM
> >   PVs for a bigger VG, in order to minimize the disks resync, so if
> >   there's a way to do that during the downtime, please do :)
> 
> This could even be done online:
> 
> 1) create two or more md's on the 2 new drives
> 2) extend the lvm VG on these new md's,
> 3) migrate all the LV's to the new md's
> 4) finally, remove the old md's from the VG
> 5) repartition the old drives as the new ones
> 6) add the old drives to the new md's to restore quadruple redundancy
> 
> However, until we're certain that the hardware is sane, we should
> probably avoid anything more sophisticated than a straight resync.

Right.

-- 
Sylvain