Re: [RGW] Space usage vastly overestimated since Octopus upgrade
Thanks, both. That’s a useful observation. I wonder what I can try to get accurate user stats. All of our users are quota-ed, so wrong users stats actually stop them from writing data. Since stats are only updated on write: I have some users who are inactive and their stats are correct. I have other users who have been actively writing. I see users who have up to 55 times the expected vs. actual size. I looped over buckets manually via the Admin Ops API and pulled the stats for all of the user’s buckets and summed these and compared that to the output from “radosgw-admin user stats" I would guess that underflowing counters could be one explanation, but there may be other things going wrong in the stats aggregation... Thanks, Liam
On Jun 30, 2020, at 6:36 AM, EDH - Manuel Rios <mriosfer@easydatahost.com> wrote:
You can ignore rgw.none details, it dont make sense today from our experience
Still dont know why dev dont cleanup bucket with those rgw.none stats...
Some of our buckets got it others new ones no.
-----Mensaje original----- De: Janne Johansson <icepic.dz@gmail.com> Enviado el: martes, 30 de junio de 2020 8:40 Para: Liam Monahan <liam@umiacs.umd.edu> CC: ceph-users <ceph-users@ceph.io> Asunto: [ceph-users] Re: [RGW] Space usage vastly overestimated since Octopus upgrade
Den mån 29 juni 2020 kl 17:27 skrev Liam Monahan <liam@umiacs.umd.edu>:
For example, here is a bucket that all of a sudden reports that it has 18446744073709551615 objects! The actual count should be around 20,000.
"rgw.none": { "size": 0, "size_actual": 0, "size_utilized": 0, "size_kb": 0, "size_kb_actual": 0, "size_kb_utilized": 0, "num_objects": 18446744073709551615 },
That number is a small negative 64bit signed value, printed as an unsigned 64bit integer. Seems like the counter underflowed.
2^64 = 18446744073709551616
-- May the most significant bit of your life be positive. _______________________________________________ ceph-users mailing list -- ceph-users@ceph.io To unsubscribe send an email to ceph-users-leave@ceph.io
participants (1)
-
Liam Monahan