Participants: Mac McCalla & Marion King (Hess Oil), Roger, Dan, Peter
- Meeting secretary: Peter
- Discussion of Hess Oil issue with DRMAAv1 interface
- SGE can put jobs into error state on abnormal conditions, which allows to continue it after manual fixing
- DRMAAv1 has no concept for this, implementation currently expresses the error jobs as still running, which makes detection difficult
- Proposal for DRMAAv2: Map such jobs to SUSPENDED state, use subState concept to express detailed error condition, transition back to RUNNING possible in the DRMAA model
- Hess Oil use case: simulation task farming, array jobs
- Discussion on job history with persistent sessions
- Agreed semantic is (still) that jobs might disappear between session usage
- On the other side, implementation can also keep history of terminated jobs in the session
- Important to not exclude this possibility in the spec layout
.- Looks good for Roger, keep it as it is.
3. Proposal for new JobInfo attributes: priority and memory consumption (check Google spreadsheet)
- Short research showed that priority is always scheduling priority for all jobs of this user
- Agreed to add priority to JobTemplate and JobInfo
- Semantic of memory consumption information in the different systems still unclear - delayed
4. Is SGE really not able to distinguish between user mode CPU time and system mode CPU time ? (check Google spreadsheet)
5. Collaborative editing of Google spreadsheet
Google spreadsheet:
Dropped due to time issues.
Best,
Peter.