I’m glad you tried it on a test system. I’ve not heard of any issues with rgw-orphan-list and I’m aware of it being used successfully on very large clusters. It is vital that you provide the correct pool for the data objects. The reason why is that the tool works by: a. Generating a list of all rados objects reachable by the bucket index b. Generating a list of all rados objects in the data pool c. Subtracting the step a. list from the step b. list So you can see that if you list the wrong pool in step b. the list you get in step c. will be misleading and removing those objects will be very problematic (to say the least). That’s in part why I said “after verification” in my previous message. Eric (he/him) > On Sep 6, 2022, at 10:34 AM, Ulrich Klein <Ulrich.Klein@xxxxxxxxxxxxxx> wrote: > > I’m not sure anymore, but I think I tried that on a test system. Afterwards I had to recreate the RGW pools and start over, so didn’t try it on a “real” system. > But I can try again in about 2 weeks. It’s dead simple to recreate the problem > (https://lists.ceph.io/hyperkitty/list/ceph-users@xxxxxxx/message/XOQXZYOWYMMQBWFXMHYDQUJ7LZZPFLSU <https://lists.ceph.io/hyperkitty/list/ceph-users@xxxxxxx/message/XOQXZYOWYMMQBWFXMHYDQUJ7LZZPFLSU>) in any Ceph version since at least Pacific > > Ciao, Uli > > >> On 6. Sep 2022, at 15:48, J. Eric Ivancich <ivancich@xxxxxxxxxx <mailto:ivancich@xxxxxxxxxx>> wrote: >> >> You could use `rgw-orphan-list` to determine rados objects that aren’t referenced by any bucket indices. Those objects could be removed after verification since this is an experimental feature. >> >> Eric >> (he/him) >> >>> On Sep 5, 2022, at 10:44 AM, Ulrich Klein <Ulrich.Klein@xxxxxxxxxxxxxx <mailto:Ulrich.Klein@xxxxxxxxxxxxxx>> wrote: >>> >>> Looks like the old problem of lost multipart upload fragments. Has been haunting me in all versions for more than a year. Haven‘t found any way of getting rid of them. >>> Even deleting the bucket seems to leave the objects in the rados pool forever. >>> >>> Ciao, Uli >>> >>>> Am 05.09.2022 um 15:19 schrieb Rok Jaklič <rjaklic@xxxxxxxxx <mailto:rjaklic@xxxxxxxxx>>: >>>> >>>> Hi, >>>> >>>> when I do: >>>> radosgw-admin user stats --uid=X --tenant=Y --sync-stats >>>> >>>> I get: >>>> { >>>> "stats": { >>>> "size": 2620347285776, >>>> "size_actual": 2620348436480, >>>> "size_utilized": 0, >>>> "size_kb": 2558932897, >>>> "size_kb_actual": 2558934020, >>>> "size_kb_utilized": 0, >>>> "num_objects": 593 >>>> }, >>>> "last_stats_sync": "2022-09-05T13:11:05.009648Z", >>>> "last_stats_update": "2022-09-05T13:11:04.955557Z" >>>> } >>>> >>>> even though I have only two empty buckets inside account. Any ideas why? >>>> >>>> Account was previously full of objects, but objects were delete few days >>>> ago. >>>> >>>> Kind regards, >>>> Rok >>>> _______________________________________________ >>>> ceph-users mailing list -- ceph-users@xxxxxxx <mailto:ceph-users@xxxxxxx> >>>> To unsubscribe send an email to ceph-users-leave@xxxxxxx <mailto:ceph-users-leave@xxxxxxx> >>> _______________________________________________ >>> ceph-users mailing list -- ceph-users@xxxxxxx <mailto:ceph-users@xxxxxxx> >>> To unsubscribe send an email to ceph-users-leave@xxxxxxx <mailto:ceph-users-leave@xxxxxxx> > _______________________________________________ ceph-users mailing list -- ceph-users@xxxxxxx To unsubscribe send an email to ceph-users-leave@xxxxxxx
 |