Cleanup #16130: Proxied operations shouldn't result in error messages if replayed - rbd - Ceph

Actions

Copy link

Cleanup #16130

closed

Proxied operations shouldn't result in error messages if replayed

Added by Jason Dillaman almost 8 years ago. Updated over 7 years ago.

Status:

Resolved

Priority:

Low

Assignee:

Vikhyat Umrao

Target version:

% Done:

Tags:

Backport:

jewel

Reviewed:

Affected Versions:

Pull request ID:

Description

For example, if the "snap rename" operation is replayed, it's possible that the second SnapshotRenameRequest state machine will encounter '-EEXISTS' when testing for the destination. This should apply to snap create, snap rename, snap remove, snap protect, snap unprotect, and rename.

# rbd bench-write -p cephfs_data --image testing2 --io-size 10240 &
# for i in {1..100}; do rbd snap rename cephfs_data/testing2@snap$i cephfs_data/testing2@snappey$i; done
2016-05-30 08:52:11.847718 7f8329ffb700 -1 librbd::SnapshotRenameRequest: encountered error: (17) File exists
2016-05-30 08:52:12.293367 7f8329ffb700 -1 librbd::SnapshotRenameRequest: encountered error: (17) File exists
2016-05-30 08:52:13.731278 7f8329ffb700 -1 librbd::SnapshotRenameRequest: encountered error: (17) File exists
2016-05-30 08:52:16.385266 7f8329ffb700 -1 librbd::SnapshotRenameRequest: encountered error: (17) File exists
2016-05-30 08:52:20.514545 7f8329ffb700 -1 librbd::SnapshotRenameRequest: encountered error: (17) File exists
2016-05-30 08:52:24.469216 7f8329ffb700 -1 librbd::SnapshotRenameRequest: encountered error: (17) File exists

Related issues 1 (0 open — 1 closed)

Actions

Copy link

Updated by Vikhyat Umrao almost 8 years ago

Status changed from New to In Progress
Assignee set to Vikhyat Umrao

Actions

Copy link

Updated by Vikhyat Umrao almost 8 years ago

@Jason Borden , I hope here you meant we should not return replay errors in Snapshot<Type>Request . For types : snap create, snap rename, snap remove, snap protect, snap unprotect, and rename if replayed.

Actions

Copy link

Updated by Vikhyat Umrao almost 8 years ago

- I have tested in latest master and I am not getting these error messages from Snapshot<Type>Request, if replayed :

./rbd snap rename rbd/testrbd@snap1 rbd/testrbd@snap1
rbd: renaming snap failed: (17) File exists

./rbd -p rbd ls -l
NAME              SIZE PARENT FMT PROT LOCK 
testrbd          1024k          2           
testrbd@snap1    1024k          2 yes       
testrbd@testsnap 1024k          2      

./rbd snap protect rbd/testrbd@snap1
rbd: protecting snap failed: (16) Device or resource busy 

$ ./rbd snap unprotect rbd/testrbd@snap1

 $ ./rbd snap unprotect rbd/testrbd@snap1
rbd: unprotecting snap failed: (22) Invalid argument

- This is my current understanding or I am missing something ?

Actions

Copy link

Updated by Jason Dillaman almost 8 years ago

@Vikhyat Umrao: yes, we should suppress the lderr log messages from the state machines if it's possible to hit the error from a duplicated op (e.g. ldout but don't lderr in SnapshotRemoveRequest if you hit -ENOENT attempting to remove a snapshot).

Actions

Copy link