Here are some quick thoughts for further diagnostics:
various times, for latency studies - time between createRun and submit to batchsub; submit to batchsub to submitted to batch; time in pending queue; running time; total latency from createRun to end of last step of a run.
can we get our hands on dataset sizes? Would be nice to plot individual sizes and totals.
I could imagine some strip charts vs time for overall monitoring, but I don't know if JAS could handle that - eg # jobs running vs time - you can find out how many are running at any given time, but can JAS keep a memory of the past times?
Will we end up with a tabbed interface as for the system tests? I can imagine a fair number of plots being generated...
2 Comments
Tony Johnson
Suggestsions for pipeline plots from Richard:
Here are some quick thoughts for further diagnostics:
Will we end up with a tabbed interface as for the system tests? I can imagine a fair number of plots being generated...
Tony Johnson
Suggestions for pipeline improvements from Bryson:
Could the PFE provide an interface that would allow suitably-authenticated users to perform rollbackFailedRun or even deleteRun operations?