]> git.apps.os.sepia.ceph.com Git - ceph-ansible.git/commit
update: reset flags before and after each osd node upgrade
authorGuillaume Abrioux <gabrioux@redhat.com>
Tue, 18 Jun 2019 08:08:48 +0000 (10:08 +0200)
committerDimitri Savineau <savineau.dimitri@gmail.com>
Fri, 8 Nov 2019 14:10:52 +0000 (09:10 -0500)
commit206ee589d6c2e01a90a6c6a6baa6a1368cb51227
treedd72a1a8c9e558bb7fced48b114502b10b119b7f
parent19a43ff26134b1a59b4cd74cc27364ab754f3d03
update: reset flags before and after each osd node upgrade

It might be possible at some point even with osd flags `noout` and
`norebalance` set the PGs states can change depending on the amount of data
written meantime. It means the check for PGs state will fail.

This commit changes the way we set those flags:
we set them before an OSD node upgrade and unset them before the PGs
state check so they can recover.

Fixes: #3961
Signed-off-by: Guillaume Abrioux <gabrioux@redhat.com>
infrastructure-playbooks/rolling_update.yml