Deployment fails on ceph-osd nodes
Affects | Status | Importance | Assigned to | Milestone | |
---|---|---|---|---|---|
Mirantis OpenStack |
Fix Released
|
High
|
Alexei Sheplyakov | ||
8.0.x |
Fix Released
|
High
|
Alexei Sheplyakov | ||
9.x |
Fix Released
|
High
|
MOS Ceph |
Bug Description
Found on Fuel 8.0 #355. Environment: perf-3 (https:/
Deployment failed, ceph-osd nodes were not deployed. Logs on the nodes show:
2015-12-28 18:57:41 ERR ceph-deploy osd prepare node-7:
When running the failed command manually:
root@node-9:~# ceph-deploy osd prepare node-9:
It fails here:
[node-9][WARNIN] DEBUG:ceph-
[node-9][WARNIN] INFO:ceph-
[node-9][WARNIN] Problem opening /dev/disk/
So, the problem is that fuel passes the journal device using "by-id" symlink (not /dev/sda3) and the current version of ceph-disk (v0.94.5) does not parse it correctly. When I run:
ceph-deploy -v osd prepare node-9:
it works.
There is a bug in ceph-disk(it should work with symlinks): http://
Diangostic snapshot will be soon.
Changed in fuel: | |
status: | New → Confirmed |
Changed in fuel: | |
assignee: | nobody → Alexei Sheplyakov (asheplyakov) |
tags: |
added: area-ceph removed: ceph |
affects: | fuel → mos |
> http:// tracker. ceph.com/ issues/ 13438
It's a different problem (caused by parted: it resolves the link and outputs its destination which confuses ceph-disk) /github. com/ceph/ ceph/blob/ v0.94.5/ src/ceph- disk#L2361- L2365 /github. com/ceph/ ceph/commit/ 0e34742b968e72a a6ce4a0c95a885d ced435b3bc /github. com/ceph/ ceph/commit/ 3bc95dfc1b88c01 e16c3df04e96acc ed777b344a /github. com/ceph/ ceph/commit/ 77ff7c3dc6dd686 1b094e5a53d329d e0802f3032
The problem is caused by extremely naive splitting of the device node path into the base device and the partition number: https:/
The bug has been silently fixed in the master branch by commits
https:/
https:/
https:/
I'm working on backporting those to hammer