]> git-server-git.apps.pok.os.sepia.ceph.com Git - ceph.git/commit
tests: add require_jewel_osds before upgrading last hammer node 13161/head
authorNathan Cutler <ncutler@suse.com>
Fri, 27 Jan 2017 21:27:18 +0000 (22:27 +0100)
committerNathan Cutler <ncutler@suse.com>
Tue, 31 Jan 2017 11:17:28 +0000 (12:17 +0100)
commit75d05809a66bee219031a7ccb64d414a2d6c8775
treee2dc9247edfe6c2ba043f5048c30309d021af8d3
parent4ef0a63e7a29eafad6dda52ea3c92846f03f268d
tests: add require_jewel_osds before upgrading last hammer node

Note: this commit was inspired by
http://github.com/ceph/ceph-qa-suite/commit/50758a4810794d265c5d36a71d1e16799251a00d

As of 10.2.4, when upgrading a cluster from hammer to jewel, after the last
node is upgraded the MON will put the cluster into HEALTH_WARN and say: "all
OSDs are running jewel or later but the 'require_jewel_osds' osdmap flag is not
set". The release notes say:

    This is a signal for the admin to do "ceph osd set require_jewel_osds" – by
    doing this, the upgrade path is complete and no more pre-Jewel OSDs may be
    added to the cluster.

Fixes: http://tracker.ceph.com/issues/18719
Signed-off-by: Nathan Cutler <ncutler@suse.com>
qa/suites/rados/singleton-nomsgr/all/lfn-upgrade-hammer.yaml
qa/suites/rados/singleton-nomsgr/all/lfn-upgrade-infernalis.yaml