Bug #48350
closed[rgw]some multipart object will become orphan objects in data pool
0%
Description
OS: CentOS 7.6
Ceph Version: 14.2.5
Procedure:
1, Use s3 browser upload a big file with multipart upload.
2, Set 'rgw_lc_debug_interval = 20' of a rgw service and set multipart upload expiration 'DaysAfterInitiation = 1'.
So the rgw lifecycle thread will run to check the multipart upload objects are expired whether or not every 20 seconds.
3, When s3 browser is still uploading some part of this file, but this file object will be expired with multipart expiration configuration.
4, Some part will upload fail and some part of this file will be left in the data pool.
And These objects will become orphan objects.
Because when the life cycle thread get part info from meta object, the uploading part does not update the omap of the multipart meta object.
How to clean the orphan obejcts in data pool?
Updated by Daniel Gryniewicz almost 3 years ago
Does this require 2 cycles, or only one? That is, if you were running normal lifecycle that runs once a day at midnight, and you started a multipart upload at 11:59, would this bug trigger? Or does it require that 2 20 second cycles happen?
Updated by Casey Bodley over 2 years ago
- Status changed from New to Can't reproduce