These are chat archives for nextflow-io/nextflow

2nd
Apr 2019
Lavanya Veeravalli
@veeravalli
Apr 02 03:55
Anyone has tried cellranger mkfastq NF+AWS? I tried running cellranger on AWS batch setup and though the commandline has --stats-dir Stats --reports-dir Reports, these two folders are not getting generated. Pipeline gets completed successfully. Fastqs are generated. Bit strange. Thanks in advance.
micans
@micans
Apr 02 09:54
@rsuchecki @stevekm good question, I'm not sure actually. It feels natural to have a single location, (and maybe with working setfacl that is easy to do (don't know!)), but we just chatted about it and for now we will have per-user cache directory indeed.
Daniel E Cook
@danielecook
Apr 02 12:28
Is there a cost to register/attend the workshop in September? Is it possible to attend just the 'camp' portion?
Evan Floden
@evanfloden
Apr 02 12:33
So far there is only the link calling for talks and workshop topics. Later today we will post the sign up link. And yes, it is possible to attend just Nextflow Camp.
Daniel E Cook
@danielecook
Apr 02 14:40
@evanfloden is there a cost to attend?
Evan Floden
@evanfloden
Apr 02 15:00
Yes, you can find the full details in the link here: https://www.crg.eu/en/event/coursescrg-nextflow-2019
Daniel E Cook
@danielecook
Apr 02 15:03
Thank you!
Sven F.
@sven1103
Apr 02 17:21
@stevekm I think, now that the weblog feature will soonish include complete workflow metadata, you should be able to easily build such monitoring app. we will do so as well at QBiC, same will be possible with @ewels idea of the monitor feature of nf-core tools. I am looking forward to all implementations :) With respect to results: as nf-core standardises the reports (and location), it is quite easy to trigger a second service via i.e. a webhook after completion and register the results. we will evaluate and most likely use StackStorm for the orchestration of events. sorry to the others, this gets a bit ot ;)
Jonathan Manning
@pinin4fjords
Apr 02 21:03

Hi! I'd like an error strategy like 'finish', but which allowed new job submissions, and only killed the main process when absolutely necessary, i.e. when no further progress can be made without the results of the failed process. Is such a thing possible?

I have some workflow runs with a very large number of processes, and just 'resuming' these runs takes forever. If individual jobs are allowed to kill the whole thing before all possible progress is made, it makes things quite difficult.

Rad Suchecki
@rsuchecki
Apr 02 22:53
@pinin4fjords errorStrategy 'ignore' ?