Actions
Bug #53885
closeddiff-iterate in fast-diff mode reports extents out of order for fancy striped images
% Done:
0%
Source:
Tags:
Backport:
octopus,pacific
Regression:
No
Severity:
3 - minor
Reviewed:
Description
$ rbd create --size 20M --stripe-unit 1M --stripe-count 3 --thick-provision foo Thick provisioning: 100% complete...done.
With fast-diff adjacent extents are reported out of order (and also aren't merged):
$ rbd diff foo Offset Length Type 0 12582912 data 12582912 8388608 data $ rbd diff --whole-object foo Offset Length Type 0 1048576 data 3145728 1048576 data 6291456 1048576 data 9437184 1048576 data 1048576 1048576 data 4194304 1048576 data 7340032 1048576 data 10485760 1048576 data 2097152 1048576 data 5242880 1048576 data 8388608 1048576 data 11534336 1048576 data 12582912 1048576 data 15728640 1048576 data 18874368 1048576 data 13631488 1048576 data 16777216 1048576 data 19922944 1048576 data 14680064 1048576 data 17825792 1048576 data
This bug goes back to the introduction of fast-diff mode in 2015.
Updated by Ilya Dryomov over 2 years ago
- Related to Bug #53784: diff-iterate reports incorrect offsets in fast-diff mode added
Updated by Ilya Dryomov about 2 years ago
- Related to Bug #55025: [test] improve fancy striping coverage added
Updated by Ilya Dryomov about 2 years ago
- Status changed from In Progress to Fix Under Review
- Pull request ID set to 45601
Updated by Ilya Dryomov about 2 years ago
- Status changed from Fix Under Review to Pending Backport
Updated by Backport Bot about 2 years ago
- Copied to Backport #55057: octopus: diff-iterate in fast-diff mode reports extents out of order for fancy striped images added
Updated by Backport Bot about 2 years ago
- Copied to Backport #55058: pacific: diff-iterate in fast-diff mode reports extents out of order for fancy striped images added
Updated by Ilya Dryomov almost 2 years ago
This bug goes back to the introduction of fast-diff mode in 2015.
It turns out that this statement is too strong. While technically true, non fast-diff modes (default and later --whole-object) behaved more or less the same through octopus. In particular, in octopus, "rbd diff foo" and "rbd diff --whole-object foo" outputs are consistent -- adjacent extents are reported out of order and aren't merged across the board there.
Based on this, I'm rejecting octopus backport as fixing non fast-diff modes in addition to fast-diff mode would be quite involved and out of scope of this issue.
Updated by Ilya Dryomov almost 2 years ago
- Status changed from Pending Backport to Resolved
Actions