I forgot that the image attachments wouldn't make it to the mailing list. Here are links to the two files mentioned below: - https://github.com/BenoitKnecht/files/blob/master/img/ceph-pool-usage.png - https://github.com/BenoitKnecht/files/blob/master/img/ceph-pool-usage-ratio.png Cheers, -- Ben On Mon, Jun 13, 2022 at 01:22:36PM +0000, Benoît Knecht wrote: > We recently noticed that the `ceph_pool_bytes_used` metric was dropping from > time to time to the level of the `ceph_pool_stored` metric, before going back > to its true value (i.e. slightly more than the `ceph_pool_stored_raw` metric). > See the attached `ceph-pool-usage.png` graph to see what I mean. > > Has anyone else observed this behavior? > > Note that this isn't just showing up in metrics, the output of `ceph df` also > contains the discrepancy. In fact, as I'm writing this I'm realizing that when > it happens, it seems to affect all the pools in the cluster at the same time. > See attached `ceph-pool-usage-ratio.png`; it shows that all pools had the > expected ratio (3 for replicated, 1.5 for erasure coded), until suddenly all > ratios dropped to 1, meaning that `ceph_pool_bytes_used` became equal to > `ceph_pool_stored`. _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |