thebestvibedev/tromcho.net
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Fix "can't get SQE, will fall out of sync with EPOLLET" when overflowing the ring

Browse Source 
OSDs shouldn't crash or hang with long iodepths anymore
This commit is contained in:
Vitaliy Filippov
2020-10-30 01:06:36 +03:00
parent 2ccb75974b
commit 23ea409081
3 changed files with 17 additions and 9 deletions
Download Patch File Download Diff File Expand all files Collapse all files
README.md
-3
View File
@@ -357,9 +357,6 @@ and calculate disk offsets almost by hand. This will be fixed in near future.
## Known Problems
- OSDs may currently crash with "can't get SQE, will fall out of sync with EPOLLET"
if you try to load them with very long iodepths because io_uring queue (ring) is limited
and OSDs don't check if it fills up.
- Object deletion requests may currently lead to 'incomplete' objects if your OSDs crash during
deletion because proper handling of object cleanup in a cluster should be "three-phase"
and it's currently not implemented. Inode removal tool currently can't handle unclean