Hi!
I started to delete multipart aborted files (about 1200 objects) and until
now it deleted 3/4 of the and the bucket usage is reduced with 2000GB from
about 3000GB in total usage!
So, it looks like 95% of the bucket size was in multimeta aborted files!
Now, the legit question is why the lifecycle policy set to delete older
files than a day is not working at all (there are still 11GB of files
visible to user, expired a month ago)! Same for aborted multipart files, it
is set to remove them after a day!
I have executed many times per day "lc process" in the last month, we have
set rgw_lc_debug_interval to something low and executed lc process. but it
ignored this bucket completly as i have in logs.
Any suggestion is welcome, as i bet we have other buckets in the same
situation.
Thank you!
Paul
On Mon, Jul 26, 2021 at 2:59 PM Paul JURCO <paul.jurco@xxxxxxxxx> wrote:
> Hi Vidushi,
> aws s3api list-object-versions shows the same files as s3cmd, so I would
> say versioning is not enabled.
> aws s3api get-bucket-versioning result is empty.
> Is there any other method to check if versioning is enabled?
> Thank you!
> Paul
>
> On Mon, Jul 26, 2021 at 2:42 PM Vidushi Mishra <vimishra@xxxxxxxxxx>
> wrote:
>
>> Hi Paul,
>>
>> Are these non-current versioned objects displayed in the bucket stats?
>> Also, the LC rule applied to the bucket can only delete/expire objects
>> for a normal bucket.
>> In the case of a versioned bucket, the LC rule applied will expire the
>> current version [create a delete-marker for every object and move the
>> object version from current to non-current, thereby reflecting the same
>> number of objects in bucket stats output ].
>>
>> Vidushi
>>
>> On Mon, Jul 26, 2021 at 4:55 PM Paul JURCO <paul.jurco@xxxxxxxxx> wrote:
>>
>>> Hi!
>>> I need some help understanding LC processing.
>>> On latest versions of octopus installed (tested with 15.2.13 and 15.2.8)
>>> we
>>> have at least one bucket which is not having the files removed when
>>> expiring.
>>> The size of the bucket reported with radosgw-admin compared with the one
>>> obtained with s3cmd is different, logs below:
>>> ~3 TB in bucket stats and 11GB from s3cmd.
>>> We tried to run manually several times the LC (lc process) but no
>>> success,
>>> even with bucket check (including with --fix --check-objects) didn't
>>> help.
>>> Configs changed are below, we have a few TB size buckets with millions of
>>> objects, 6h default processing LC time was always not enough:
>>> rgw_lc_debug_interval = 28800
>>> rgw_lifecycle_work_time = 00:00-23:59
>>> rgw_lc_max_worker = 5
>>> rgw_lc_max_wp_worker = 9
>>> rgw_enable_lc_threads = true
>>>
>>> Status of LC is always complete:
>>> {
>>> "bucket":
>>>
>>> ":feeds-bucket-dev-dc418787:3ccb869f-b0f4-4fb9-a8d7-ecf5f5e18f33.37270170.140",
>>> "started": "Mon, 26 Jul 2021 07:10:14 GMT",
>>> "status": "COMPLETE"
>>> },
>>> It looks like versioning is not supported yet, and with aws-cli I did not
>>> get any response for 's3api get-bucket-versioning'.
>>> So, what to do?
>>>
>>> $ s3cmd -c .s3cfg-feeds-bucket-dev-dc418787 du
>>> s3://feeds-bucket-dev-dc418787
>>> *12725854360 *192 objects s3://feeds-bucket-dev-dc418787/
>>> (192 files, 11GB)
>>>
>>> Output of 'radosgw-admin lc get' and 'bucket stats' is attached.
>>> Thank you for any suggestion!
>>>
>>> Also, we found that the LC is removing files earlier than set for another
>>> bucket, 6h after the file was added instead of 31 days.
>>>
>>> Paul
>>> _______________________________________________
>>> ceph-users mailing list -- ceph-users@xxxxxxx
>>> To unsubscribe send an email to ceph-users-leave@xxxxxxx
>>>
>>
_______________________________________________
ceph-users mailing list -- ceph-users@xxxxxxx
To unsubscribe send an email to ceph-users-leave@xxxxxxx
