Feature #602
closed
mds: handle corrupt/missing journals
Added by Sage Weil over 13 years ago.
Updated over 7 years ago.
Description
This probably means
- shutting down current instances, resetting cluster membership
- throwing out journals (or maybe not; but that's a separate feature)
- restarting cluster in single-mds, maybe with repair mode flag
- Tracker changed from Tasks to Feature
- Source changed from 0 to 3
- Tracker changed from Tasks to Feature
- Target version changed from 12 to v0.25
- Parent task deleted (
#86)
- Translation missing: en.field_position set to 7
- % Done changed from 0 to 60
see mds_reset branch.
Still need a process to reset the individual mds instances. This should:
- advance the journal header info to a new, unused region.
- do, or journal a single event that (on replay) will,
- update the mds inotable to avoid reusing ino's.
- zero out the mds session table.
- Status changed from New to Resolved
- % Done changed from 60 to 100
no repair flag, but journal discard and cluster reset are in place. not especially useful yet without the actual repair, though.
commit:548f329f03a640e292cd6ce762ffbacde3017e90
- Project changed from Ceph to CephFS
- Category deleted (
1)
- Target version deleted (
v0.25)
Bulk updating project=ceph category=mds bugs so that I can remove the MDS category from the Ceph project to avoid confusion.
Also available in: Atom
PDF