It sounds like you've completely lost the PG, but *may* still have enough object data to reconstruct them. However, there's no tooling to make this easy. You'll need to build your own scripts (and probably binaries) to reconstruct the original data from all the input shards, and force-create the PGs to get them back up. -Greg
On Thu, May 31, 2018 at 12:00 AM Monis Monther <[email protected]> wrote: > Hi, > > We have an EC pool that is 3+1 (luminous 12.2.0 using filestore), 2 OSDs > crashed and some PGs are now incomplete because we are left only with 2 out > of 4. The failed OSDs have the two other chunks of the PGs intact but they > are missing the omap directory and other files. We replaced the failed > disks with new ones maintaining the OSD id and crush map. Now we are trying > to import back the lost chunks of the PG. > > We tried to export the PG using ceph-objectstore-tool, but it fails to > export. > > We also tried to copy (including xattrs) the dir directly to the new OSDs, > however the OSD did not start. > > Is there a way to recover those PGs? > > > -- > Best Regards > Monis > _______________________________________________ > ceph-users mailing list > [email protected] > http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com >
_______________________________________________ ceph-users mailing list [email protected] http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
