On 04/16/2018 11:14 AM, Yehuda Sadeh-Weinraub wrote:
> Should have probably used "${bucket_marker_id}_" here
I double checked the results of "fgrep $bucket_marker_id" and "fgrep
$[bucket_marker_id}_" and there were no rados objects that got deleted due to
the loose matching of the first one but have taken a note in case I'll need to
do similar again.
>
> I think that unless your grep there found entries on other bucket
> instances (due to bucket marker id happen to be a prefix of some other
> bucket's marker id) you're probably safe. Not sure about that crash,
> if you have logs feel free to share them. It's not clear to me if the
> logs above were transient, or whether you still see them. If you still
> see them then you'd want to look into the culprit.
The crash hasn't happened again so I guess it was just a coincidence.
I guess I better disable dynamic sharding for now and set up a separate cluster
to load- and QA-test this, see if I can find anything that would explain why
the resharding process was blocked indefinitely in the first place.
_______________________________________________
ceph-users mailing list
ceph-users@lists.ceph.com
http://lists.ceph.com/listinfo.cgi/ceph-users-ceph.com
