ceph-osd unnecessarily chowns /var/lib/ceph firefly->hammer
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Ceph OSD Charm |
Fix Released
|
High
|
Billy Olsen | ||
ceph-osd (Juju Charms Collection) |
Invalid
|
High
|
Billy Olsen |
Bug Description
When upgrading from Firefly to Hammer, the charm unnecessarily recursively chowns the /var/lib/ceph filesystems. The charm does not need to perform the chown unless the upgrade is to a level of code which runs the ceph daemons as the ceph user (introduced in Infernalis).
In a production environment with numerous OSDs and lots of objects stored, this recursive chown takes a significant time since it crawls filesystems for all OSDs.
The code should intelligently chown the filesystems depending on the upgrade path.
This can be seen in the code [0] for master and [1] for the 16.10 charms, in which the chown is occurring regardless of whether it is necessary or not.
[0] https:/
[1] https:/
Since there are no precautions to prevent the OSD from getting kicked out (e.g. noout is not set), this causes the entire node's worth of OSDs to be kicked out of the Ceph cluster causing data movement and recovery.
Changed in ceph-osd (Juju Charms Collection): | |
milestone: | none → 17.01 |
Changed in ceph-osd (Juju Charms Collection): | |
status: | Confirmed → In Progress |
Changed in charm-ceph-osd: | |
assignee: | nobody → Billy Olsen (billy-olsen) |
importance: | Undecided → High |
status: | New → In Progress |
Changed in ceph-osd (Juju Charms Collection): | |
status: | In Progress → Invalid |
This was fixed in rev 242 of the charm in the charmstore.