Actions
Bug #62720
openmds: identify selinux relabelling and generate health warning
Status:
In Progress
Priority:
High
Assignee:
Category:
Administration/Usability
Target version:
% Done:
0%
Source:
Tags:
Backport:
reef,quincy
Regression:
No
Severity:
3 - minor
Reviewed:
Affected Versions:
ceph-qa-suite:
Component(FS):
MDS
Labels (FS):
Pull request ID:
Crash signature (v1):
Crash signature (v2):
Description
This request has come up from folks in the field. A recursive relabel on a file system brings the mds down to its knees - not an mds issue as such, its a badly designed upper layer that applies `security.selinux` extended attribute touching all inodes in a file system before handing the file system path for mounting (subvolume users). Using a perf counter to track the relabel and emitting a warning when it exceeds a threshold would ease of some burden for cluster admins (who currently have to either analyse debug log files to be certain that the slowness is due to an ongoing selinux relabel).
Updated by Venky Shankar 5 months ago
- Backport changed from reef,quincy,pacific to reef,quincy
Updated by Venky Shankar 3 months ago
- Assignee changed from Manish Yathnalli to Christopher Hoffman
Chris, please take this one whenever you get some time off from the fscrypt work :)
Actions