Mimic clients can't talk to mixed M+N Cluster
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Ceph |
Unknown
|
Unknown
|
|||
ceph (Ubuntu) |
Confirmed
|
High
|
Unassigned |
Bug Description
When trying to upgrade from Mimic to Nautilus, the clients on the Mimic nodes fail to talk to the monitor cluster as soon as the first unit is upgraded:
# ceph -s
Traceback (most recent call last):
File "/usr/bin/ceph", line 1241, in <module>
retval = main()
File "/usr/bin/ceph", line 1165, in main
sigdict = parse_json_
File "/usr/lib/
cmd['sig'] = parse_funcsig(
File "/usr/lib/
raise JsonFormat(s)
ceph_argparse.
The monitor cluster is actually still functional according to that same command on the newly Nautilus unit:
$ sudo ceph -s
cluster:
id: 2734983c-
health: HEALTH_OK
services:
mon: 3 daemons, quorum juju-a792f2-
mgr: juju-a792f2-
osd: 3 osds: 3 up, 3 in
rgw: 1 daemon active (juju-a792f2-
data:
pools: 15 pools, 46 pgs
objects: 187 objects, 1.1 KiB
usage: 3.0 GiB used, 27 GiB / 30 GiB avail
pgs: 46 active+clean
Changed in ceph (Ubuntu): | |
status: | New → Confirmed |
no longer affects: | ceph |
Changed in ceph (Ubuntu): | |
importance: | Undecided → High |
Configuring the `ceph` command to run on an upgraded monitor unit only (`ceph -m $MON_IP`) causes the command to work successfully