PR status #183
Description
Since our PRs are so stagnent, I made a summary of them below to hopefully help maintainers decide what to do. Or maybe encourage more people to take a look.
Six should be used right away, some are old. At least three can be closed.
- travis: Attempt to fix CI #180 (fix CI) Can be used immediately.
- Introduce unknown job status to handle communication problem with resource manager #179 (unknown job status) some comments but can then be used. Needs changelog entry.
- When checking the job status crashes, do not force empty status, keep… #178 duplicate of Introduce unknown job status to handle communication problem with resource manager #179, Introduce unknown job status to handle communication problem with resource manager #179 should be used instead. Close when Introduce unknown job status to handle communication problem with resource manager #179 accepted.
- Fail hard on first error in batch script #177 (job script set -e) - makes sense, someone either needs to do it and accept some people's use cases will change, or accept that we can never change it. Needs changelog entry.
- add 'gres' option to SlurmSpawner #170 (slurm --gres option) Should be accepted if we want more options, instead of the generic req_options. Needs changelog entry, I already added one but it is out of date since the release.
- Update batchspawner.py to use --chdir instead of --workdir #165 Can be used, I have a local changelog entry I couldn't push, but of course have to update since it's so old, and there has been a release.
- Grab environment variables needed for grid engine #137 (fix grid engine) can probably be used. more importantly, we should do some community building to bring these contributors in.
- Initial support for OAR batch scheduler #136 (OAR) I would probably reject this, since the OAR scheduler seems sort of insecure, and make it some sort of community recipe. We can help people do it, but shouldn't make it easy to do insecure things out of the box.
- Environment handling refactoring. #119 (environment refactoring) This was a draft test of mine, no point to work on it until we have a way to accept new ideas.
- Select random port from a range (new implementation) #114 Now a bit obsolete, with select port on remote side. Needs later thought, after we have a way to accept new ideas.
- [WIP, demo only] command to terminate jobs before killing them #75 (sigterm before sigkill) Not really needed with slurm, but could be useful other ways. Not something to think about, until we have solved the general PR problem.