Closed LudvigOlsen closed 3 months ago
That sounds weird. I haven't heard of any other instances of this happening. It could be many many things not related to gwf at all (the filesystem, Slurm reporting intermittent states, jobs not being completely finished, but being counted as "completed" by gwf for a short while).
Have you experienced this since you reported it?
Hi Dan, I haven't experienced it since no. Not sure what went wrong. I see how it could be a problem in slurm and not GWF.
I have recently started using the following instead of gwf status
since it returns instantly (gwf status
takes 45min with my giant workflow :-) ):
alias sqs='squeue -u $USER -o "%.10i %.9P %.8j %.8u %.2t %.10M %.6D %.6R" | awk '\''
NR > 1 {
state = $5
if (state == "R") state = "Running"
else if (state == "PD") state = "Pending"
else if (state == "CG") state = "Completing"
else if (state == "CD") state = "Completed"
else if (state == "F") state = "Failed"
else if (state == "TO") state = "Timeout"
# Add more state mappings as needed
count[state]++
}
END {
for (status in count) print status, count[status]
}'\'
I'll close this for now then :-)
(gwf status
needs to ask for metadata on every single file in your workflow to determine if everything is up-to-date, while squeue
just gives you the list of jobs and their status).
Last night, I ran the gwf status summary twice in a row in my workflow. I didn't change anything in-between. Jobs that were marked as completed in the first summary were somehow marked as shouldrun in the second summary.
gwfss
is just an alias for making the summary with some timestamps around it. So:alias gwfss="echo '----------------------------'; echo 'gwf status'; date; gwf status -f summary; date; echo '----------------------------'"
.I am not sure how to reproduce this. It seems kind of random since I didn't change any files or anything.