semantics of arguments to `example7` and `example10`
Closed this issue · 4 comments
I only caught this when running them back-to-back, but when you run bookkeeper.py 5 in example7, it actually runs 10 jobs (5 of each). When you run any of the scripts in example 10 with 10 as the argument, then 10 jobs get run (5 of each). I think the semantics of bookkeeper should be changed (if possible) to match those of the waits in example10.
Thanks for pointing this out @SteVwonder! I agree, the semantics for this could definitely be made clearer.
On another note on example7: what version of Flux are you running this example against? I just ran this example against flux-core v0.14 and it looks like job_state_cb() never gets called. Here's my output:
bash-4.2$ ./bookkeeper.py 3
187133067264
187468611584
187888041984
188324249600
188726902784
189230219264
bookkeeper: all jobs submitted
bookkeeper: waiting until all jobs complete
bookkeeper: all jobs completed
However, the jobs will run for their set time (120 seconds in this case):
bash-4.2$ flux jobs
JOBID USER NAME STATE NTASKS TIME
189230219264 moussa1 io-forward RUN 3 2.8s
188726902784 moussa1 compute.py RUN 6 2.9s
188324249600 moussa1 io-forward RUN 3 2.9s
187888041984 moussa1 compute.py RUN 6 2.9s
187468611584 moussa1 io-forward RUN 3 2.9s
187133067264 moussa1 compute.py RUN 6 3s
I know that Frank D made some changes to example 7 to make it work under the latest version of Flux. Here is the final script we ended up with for the tutorial: https://github.com/flux-framework/Tutorials/pull/3/files#diff-3c1bff2672c3ffd8cba68cb8f102cdce
Does that fix the issue?
Just ran it - still got that same output listed as above. Maybe it could have something to do with the Flux version I am running against? I am running it against a locally-installed version of Flux. I can open a separate issue for it.