It looks like these messages are related to the config variable rgw_bucket_quota_soft_threshold, which defaults to 0.95. I dug through the git history and found this was added in a 2013 commit https://github.com/ceph/ceph/commit/14eabd4aa7b8a2e2c0c43fe7f877ed2171277526. I guess the reasoning there is that, once a bucket is close to hitting its quota, we want our quota checks to be 'exact' instead of using the cache. But these quota checks can be extremely expensive for sharded buckets, and the checks aren't atomic with the writes anyway. The change long predates dynamic resharding, and I don't think it's reasonable anymore. I'd support reverting that commit entirely. What does everyone else think? On Mon, Jul 6, 2020 at 6:44 AM Seena Fallah <seenafallah@xxxxxxxxx> wrote: > > Hi all. > > I'm facing this log on my rgw instances and this seems the reason that > I have so high iops on my buckets.index pool. > > 2020-07-04 18:15:08.472 7f15b37fa700 20 quota: can't use cached stats, > exceeded soft threshold (size): 515396075520 >= 489626271744 > > Can someone help me on this? > > Thanks. > _______________________________________________ > Dev mailing list -- dev@xxxxxxx > To unsubscribe send an email to dev-leave@xxxxxxx > _______________________________________________ Dev mailing list -- dev@xxxxxxx To unsubscribe send an email to dev-leave@xxxxxxx
