I've submitted a few thousand jobs which I now realise may be using the wrong data.
How can I terminate all of these as quickly as possible?
Is there a better way than looping through job IDs calling dx terminate for each? This seems fairly slow.
Looping through all job is the fastest way. I think you can do it quickly though.
1) gather all job ID you need to terminate. If you have subjob, you can consider gather just root job since subjub would be terminated anyway if the main job is terminated. Or if you didn't expect to have ton of subjob, you can do dx find jobs --all-jobs which would cut the processing time by avoid recontructing job tree. You can narrow down search scope for just status and type of job you want too.
2) from a file containing job-id, you can split it into a few of them.
3) for each split chunk, use a new terminal to loop through and kill those jobs.
Comments
1 comment
Looping through all job is the fastest way. I think you can do it quickly though.
1) gather all job ID you need to terminate. If you have subjob, you can consider gather just root job since subjub would be terminated anyway if the main job is terminated. Or if you didn't expect to have ton of subjob, you can do dx find jobs --all-jobs which would cut the processing time by avoid recontructing job tree. You can narrow down search scope for just status and type of job you want too.
2) from a file containing job-id, you can split it into a few of them.
3) for each split chunk, use a new terminal to loop through and kill those jobs.
Please sign in to leave a comment.