Bug #53457
openStale bucket index entry remains after object deletion
0%
Description
I found a special sequence may cause this new situation
1.requests an object be deleted,prepares the op, adding pending ops to the bucket dir entry
2.list bucket
3.sees that there are pending operations on bucket dir entry, and calls check_disk_state
4.check_disk_state sees that the object still exists,so it sends CEPH_RGW_UPDATE to bucket index osd
5.if aio_operate dir_suggest_changes find the pending expired,erase pending
6.deletes the object
7.index complete
step7 found pending not exist,so not remove index
Updated by Casey Bodley over 2 years ago
- Status changed from New to Need More Info
- Tags set to dir_suggest
in step 5, the pending operations shouldn't expire for 120 seconds. did the delete really take that long to process?
Updated by zhong wang over 2 years ago
I found some stale bucket indexs in my project. Otherwise,I can't find any other reason for this problem
Updated by Jonas Nemeikšis 7 months ago
Zhong, maybe found more reasons? What's version?