Actions
Feature #19703
openIf there is one FULL OSD, All IO the the cluster will be halted(read only).
Status:
New
Priority:
Normal
Assignee:
-
Category:
-
Target version:
-
% Done:
0%
Source:
Tags:
Backport:
Reviewed:
Affected Versions:
Pull request ID:
Description
If there is one FULL OSD, All IO the the cluster will be halted(read only). This is not good for a large scale multi-pool deployed Ceph.
Consider such a fine granularity policy: If the FULL OSD belongs to Pool-A, then all IOs to Pool-A will be affected, IOs to Pool-B, Pool-C… will continue as usual. Is this a more reasonable policy and worthy implementing it?
Actions