Merge lp://staging/~stylesen/lava-scheduler/multinode into lp://staging/lava-scheduler/multinode
Proposed by
Senthil Kumaran S
| Status: | Merged | ||||
|---|---|---|---|---|---|
| Approved by: | Neil Williams | ||||
| Approved revision: | no longer in the source branch. | ||||
| Merged at revision: | 249 | ||||
| Proposed branch: | lp://staging/~stylesen/lava-scheduler/multinode | ||||
| Merge into: | lp://staging/lava-scheduler/multinode | ||||
| Diff against target: |
347 lines (+299/-7) 3 files modified
lava_scheduler_app/migrations/0031_auto__add_field_testjob_target_group.py (+161/-0) lava_scheduler_app/models.py (+52/-7) lava_scheduler_daemon/utils.py (+86/-0) |
||||
| To merge this branch: | bzr merge lp://staging/~stylesen/lava-scheduler/multinode | ||||
| Related bugs: |
|
| Reviewer | Review Type | Date Requested | Status |
|---|---|---|---|
| Neil Williams | Approve | ||
|
Review via email:
|
|||
Description of the change
Support for submitting multinode jobs to the scheduler and have sub-ids and unique target-groups.
To post a comment you must log in.
Preview Diff
[H/L] Next/Prev Comment, [J/K] Next/Prev File, [N/P] Next/Prev Hunk
OK, there are a few debug bits left in the split_multi_job function. The group_json doesn't need to know the hostname once the code from translate.py is migrated into the scheduler. Firstly because that will be the hostname of the machine running the scheduler daemon, not the machine running the GroupDispatcher. During debugging, translate.py was run on the same machine as the dispatcher. As the comment notes, the only thing the GroupDispatcher needs from group_json is the group_dispatcher = true flag and the port number. (In some ways, group_json is possibly redundant and the port number could be passed as an argument to __init__.)
However, this is harmless, so I'm happy to approve this merge, we will need to remove the hostname call (and therefore the import of hostname from socket) before the branch is finally merged.