]> git-server-git.apps.pok.os.sepia.ceph.com Git - ceph.git/commit
mon/FSCommands: add 'recover' flag in `fs new` command 42763/head
authorRamana Raja <rraja@redhat.com>
Wed, 11 Aug 2021 20:34:47 +0000 (16:34 -0400)
committerRamana Raja <rraja@redhat.com>
Mon, 13 Sep 2021 04:15:39 +0000 (00:15 -0400)
commit67bb13859a248a03059507ec52b4f551db0a72c9
tree5bdb352d697903d2d76715fea6dff5a3a26537e9
parentccccbff70bc149074bff0aba5230ac80138aed92
mon/FSCommands: add 'recover' flag in `fs new` command

Currently, to recover a file system after recovering monitor store, you
need to stop all the MDSs; create FSMap with defaults using `fs new`
command; execute `fs reset` command to get the file system's rank 0 into
existing but failed state; and then restart MDSs.

Add 'recover' flag to the `fs new` command that sets the file system's
rank 0 to existing but failed state, and sets the file system's
'joinable' setting to False. Using the `fs new` command with 'recover'
flag gets rid of the steps to stop all the MDSs and execute `fs reset`
command when recovering the file system after recoving monitor store.

Fixes: https://tracker.ceph.com/issues/51716
Signed-off-by: Ramana Raja <rraja@redhat.com>
doc/cephfs/administration.rst
doc/cephfs/recover-fs-after-mon-store-loss.rst
qa/suites/fs/functional/tasks/recovery-fs.yaml [new file with mode: 0644]
qa/tasks/cephfs/test_recovery_fs.py [new file with mode: 0644]
src/mds/FSMap.cc
src/mds/FSMap.h
src/mon/FSCommands.cc
src/mon/MonCommands.h