These are chat archives for dereneaton/ipyrad

Apr 2016
Isaac Overcast
Apr 10 2016 19:43
I like the new system so far, but my biggest concern is that if jobs fail the lbview will keep trying all the other queued jobs, this is where ImpossibleDependency is coming from, a job that is after a failed job can't ever be run, so it's null'd, the result is a cascade of failures up to lbview.apply(time. which succeeds, so then the whole step looks like it succeeds. I've been adding some error handling to catch exceptions, but we should figure out a good way to deal with them so step3 actually reports failure on failure... still workin on it. Doesn't help that i've been layed out sick this weekend, hard to code when the fever keeps your brainstack at about 3 items ;p
Isaac Overcast
Apr 10 2016 20:19
denovo +/- reference works in apply3 now
testing on real data instead of sim
Deren Eaton
Apr 10 2016 22:00
According to the docs there is a flag similar to 'after' called 'failed' which can be used to tell ipyparallel what to do when an exception occurs. We should submit jobs with the failed flag that cleans up and prints async object metadata on failure. I haven't had time to figure out how to do it yet, tho.