Feature #2574
closed
crowbar: use data disks automatically, journal inside data directory
Added by Anonymous almost 12 years ago.
Updated almost 12 years ago.
Description
Crowbar sets node['crowbar']['disks'] to an array of disks. First one is used for the OS, and disk['usage'] is set to 'OS'.
Use all disks in node['crowbar']['disks'] that don't have 'usage' set, set their usage to 'ceph-osd'. Use the $osd_data_dir/journal file as journal, for now.
Warning: if Swift is installed on the same node, it'll try to use the data disks too! Make barclamp-ceph refuse to use those nodes?
- Translation missing: en.field_position set to 3
- Translation missing: en.field_position deleted (
3)
- Translation missing: en.field_position set to 1
- Translation missing: en.field_story_points set to 3
- Translation missing: en.field_position deleted (
2)
- Translation missing: en.field_position set to 2
- Translation missing: en.field_position deleted (
14)
- Translation missing: en.field_position set to 1
- Translation missing: en.field_position deleted (
6)
- Translation missing: en.field_position set to 2
- Project changed from Ceph to devops
- Category deleted (
chef)
- Target version set to v0.50
- Translation missing: en.field_position deleted (
2)
- Translation missing: en.field_position set to 84
- Status changed from New to In Progress
- Assignee set to JuanJose Galvez
- Target version changed from v0.50 to v0.51
- Translation missing: en.field_position deleted (
91)
- Translation missing: en.field_position set to 2
- Translation missing: en.field_position deleted (
3)
- Translation missing: en.field_position set to 1
The most recent pull request for the cookbook has been tested by Tyler and myself. I've setup the following situations during my testing:
3 mons with 1 osd.
3 mons with 3 osd on the mon nodes.
3 mons with 6 osd, none on the mon nodes.
3 mons with 9 osd, three of those on the mon nodes.
I also setup nodes using swift, verified that those disks are skipped.
All of these end with a working ceph cluster.
- Status changed from In Progress to Resolved
There were bugs and the history was wrecked by github pull requests again, so I redid some commits, but this functionality is now in -- except for the unrelated bugfix that did weird things with df that I want to understand better before merging.
Also available in: Atom
PDF