If I lose an OST and have all the data from the entire cluster backed up, how do I recover it? (not a backup of the OST, but a backup of the contents of the cluster) I know there is data in the MGT that indicates where the data is stored, but how can I put it back in the cluster and replace the OST? If I set stripping to 1 (no file striping) does that ease the problem? Thanks! -- Andrew -------------- next part -------------- An HTML attachment was scrubbed... URL: http://mail.clusterfs.com/pipermail/lustre-discuss/attachments/20070924/74e078dc/attachment.html
We had to do this once using Lustre 1.4. The details will have changed slightly
for 1.6 no doubt but the basics are:
1) replace/reformat the bad RAID and mount/add to LustreFS
* you will now have lots of "0" byte files on your filesystem
2) generate a list of all files that were on the OST using "lfs find --obd
ost1d_UUID /mnt/lustre"
3) feed this list to "cpio" or similar to copy the files from your
backup over the "0-byte" files - they will be written to the empty OST
You''ll need to flesh out the details from the manual depending on
Lustre 1.4/1.6. With 1.6 you probably have to set the ost index manually to
match the old one in the mkfs.lustre command to stop it just adding a new OST.
Regards,
Daire
----- "Andrew Lundgren" <Andrew.Lundgren at Level3.com>
wrote:> _______________________________________________
> Lustre-discuss mailing list
> Lustre-discuss at clusterfs.com
> https://mail.clusterfs.com/mailman/listinfo/lustre-discuss
Thanks for your reply! I will look into the details on 1.6. -- Andrew> -----Original Message----- > From: Daire Byrne [mailto:Daire.Byrne at framestore-cfc.com] > Sent: Tuesday, September 25, 2007 4:00 AM > To: Lundgren, Andrew > Cc: Lustre-discuss at clusterfs.com > Subject: Re: [Lustre-discuss] Recovery of a failed OST. > > > We had to do this once using Lustre 1.4. The details will > have changed slightly for 1.6 no doubt but the basics are: > > 1) replace/reformat the bad RAID and mount/add to LustreFS > * you will now have lots of "0" byte files on your filesystem > 2) generate a list of all files that were on the OST using > "lfs find --obd ost1d_UUID /mnt/lustre" > 3) feed this list to "cpio" or similar to copy the files from > your backup over the "0-byte" files - they will be written to > the empty OST > > You''ll need to flesh out the details from the manual > depending on Lustre 1.4/1.6. With 1.6 you probably have to > set the ost index manually to match the old one in the > mkfs.lustre command to stop it just adding a new OST. > > Regards, > > Daire > > > ----- "Andrew Lundgren" <Andrew.Lundgren at Level3.com> wrote: > > _______________________________________________ > > Lustre-discuss mailing list > > Lustre-discuss at clusterfs.com > > https://mail.clusterfs.com/mailman/listinfo/lustre-discuss > >