IRC channel logs

2020-11-30.log

back to list of logs

<efraim>almost there on a slurm set of services! slurmd+slurmctld+slurmdbd fail to start, but when started manually with the flags I gave them in their services I get 'srun --label --ntasks=1 hostname' to print the hostname
<civodul>efraim: woohoo, neat!
<civodul>getting very close
<efraim>I'm getting around the config files by using a simple-service for etc-service-type and plopping the config files in place
<efraim>there's so many options, and some change by version, I don't see doing much past "and here's your text field, fill it up!"
<civodul>heheh
<rekado_>zimoun: I’m officially back to work today, so I’d like to finish the R upgrades soon. Is there some work-in-progress that you have been working on?
<rekado_>do you happen to know if there are any security certifications that Guix could undergo to get a marketable stamp of approval?
<rekado_>our research group has been thinking about advertising source-to-binary transparency and container provenance; this is pretty important to become confident about suitability for data privacy sensitive applications
<civodul>hi rekado_!
<civodul>wb to work (?) :-)
<civodul>there was a discussion with the usual suspects in .fr about making a manifesto on reproducibility and what it entails
<civodul>that's not really a "marketable stamp", but it could be a step in that direction
<rekado_>yes, any “gimmick” would help, I think
<rekado_>just like the very mention of research institutes on the Guix HPC website helps lend a little bit of authority to the project itself
<rekado_>it’s been frustrating to me how reliably the community keeps falling back onto the lowest common denominator: Conda and Docker
<rekado_>our message has gotten through enough to make those suggesting these things explain themselves, but it’s not enough
<rekado_>(like, you’ll hear people say “I know that it’s not the most advanced in reproducibility, etc, but it’s good enough for all cases I care about”)
*rekado_ applies Roel’s upgrade to the rstudio and rstudio-server packages
<civodul>yeah
<civodul>i think every concrete demonstration helps
<civodul>like you did for PiGx
<civodul>also debunking myths (currently i'm focusing again on the perceived "performance issues")
<civodul>BTW, did you hear from birdsite?
<efraim>I think the log file killing my slurmd process :/
<efraim>when the logfile exists starting slurmd removes it and I get a 'In procedure fport_write: Bad file descriptor' error
<efraim>always fun when you have to carefully parse man pages to figure it out
<efraim>Run slurmd in the foreground. Error and debug messages will be copied to stderr.
<efraim>KEY WORD COPY
<efraim>slurmd and shepherd both tried to take control of the log file
<rekado_>civodul: no news from birdsite support
<rekado_>I wrote them once more but I only received their automatic email replies
<rekado_>nothing that a human wrote
<rekado_>my colleague pointed me to this: https://www.enisa.europa.eu/
<rekado_>in particular this: https://www.enisa.europa.eu/publications/healthcare-certification
<civodul>hmm ok
<civodul>i guess we should get in touch with people in charge of such initiatives
<civodul>the ideal for me would be to talk to the ACM re https://www.acm.org/publications/policies/artifact-review-badging
<rekado_>this currently only applies to ACM publications; do you mean to talk to them about something wider in scope?
<civodul>we could talk to the ACM about ways they could recommend functional deployment tools like Guix as part of their reproducible research policy
<civodul>it'd "only" apply to ACM publications, but that's already a lot
<civodul>similarly, if there are societies in other domaines that have a RR policy, we should talk to them
<civodul>BTW, i encourage y'all to join https://reproducible-research.inria.fr/
<civodul>it's an open discussion forum about RR
<efraim>good news! I'm bad at match-lambda. forgot to include the skipped variables. just have a remaining issue with slurmstepd and changing directories
<civodul>heh :-)
***nckx is now known as ncksneekx
***ncksneekx is now known as nckx