search

hammerlab / biokepi

Bioinformatics Ketrew Pipelines
Apache License 2.0
27 stars 4 forks source link

Use local tmp dir for NetMHC family #465

Closed armish closed 7 years ago

armish commented 7 years ago

Since the one on the mounted folder seems to be causing issues from time to time (reported by @julia326).

smondet commented 7 years ago

I don't think /tmp should be set in stone there without being configurable in some way, very often biokepi nodes run on machines that have very little local storage (and bfx tools that use temp-files usually, it's because tmp-data doesn't fit in memory, e.g. samtools sort).

julia326 commented 7 years ago

With NetMHC tools though, it seems like they don't play nice with anything but local storage - what do you think we should do in this case?

smondet commented 7 years ago

@julia326 @armish I just think it should be somewhat configurable, from the biokepi-machine for example (even if /tmp is the default, it's just that we've been bitten by these things quite a few times).

Also, is there an issue reported somewhere about NetMHC not behaving posixly?

ihodes commented 7 years ago

I believe this works (solves the tmp dir issue + seems to have allowed NetMHC to run correctly!)

armish commented 7 years ago

Phew - glad this solved your issues!

@julia326: I don't think we have an issue describing this problem but I am not sure where would be the best place to put this information. Maybe mentioning this in the bundle README will help others to not hit this in the future (and most importantly without noticing)?

In terms of allowing this to be customized: I do agree that this should be customizable and definitely on the biokepi level. I am not sure whether we should surface yet another ENV variable to the epidisco and alike. @smondet @ihodes: what do you think?

julia326 commented 7 years ago

@armish I can put an issue into biokepi, that seems like the best place for it. Can also add a note to the netmhc-bundle README, though luckily in this case the netMHC error was informative enough that it became pretty clear what the problem is.

--> just filed https://github.com/hammerlab/biokepi/issues/467 to track

armish commented 7 years ago

@smondet: just came here to say this is ready for your review and 😍