#general
- h
happy-kitchen-89482
11/09/2015, 6:47 PMNo single dramatic thing - m
mammoth-pharmacist-54215
11/09/2015, 6:47 PMdid you have to go through and do a cleanup of CI scripts or anything like that? - h
happy-kitchen-89482
11/09/2015, 6:48 PMCIs do start with a clean - h
happy-kitchen-89482
11/09/2015, 6:48 PMiirc - h
happy-kitchen-89482
11/09/2015, 6:48 PMbecause we artifact-cache EVERYTHING, so a clean build isn’t even a big deal for us. - b
bored-art-40741
11/09/2015, 6:48 PM@rough-minister-58256: That is our experience as well. We used to have a
that is effectively the./pants scrub
being discussed herenuke - b
bored-art-40741
11/09/2015, 6:49 PMNot only did it not every properly capture everything that could be going wrong (lots of things are sideband but related to pants in the end), but it took FOREVER to train users to not use it whenever they saw something even slightly funny - b
bored-art-40741
11/09/2015, 6:49 PMIn the end, I just removed it - h
happy-kitchen-89482
11/09/2015, 6:49 PMHmm - r
rough-minister-58256
11/09/2015, 6:49 PMyeah. - h
happy-kitchen-89482
11/09/2015, 6:49 PMso maybe this is an argument against a
?./pants nuke - b
bored-art-40741
11/09/2015, 6:49 PMI would argue against it - h
happy-kitchen-89482
11/09/2015, 6:49 PMOr at least, if Twitter really need it, it can be a custom task or an external script. - b
bored-art-40741
11/09/2015, 6:49 PMI would prefer to see
and./pants invalidate
takeclean
argumentsgoal.task - b
bored-art-40741
11/09/2015, 6:50 PMYeah, your nuke should be internal, if anything. - h
happy-kitchen-89482
11/09/2015, 6:50 PMYeah, with the new engine, and the total isolation of products, clean could be properly selective. - r
rough-minister-58256
11/09/2015, 6:50 PMmaybe it could throw a huge warning banner indicating that its not a good idea to run often, etc - b
bored-art-40741
11/09/2015, 6:50 PMThere's always too much sideband stuff in a given install to expect a generic nuke to ever do the right thing - b
bored-art-40741
11/09/2015, 6:50 PMimo nuke would be "spin up a fresh docker image and run your command again" - b
bored-art-40741
11/09/2015, 6:51 PMAnything else falls short while charging the user a lot of time rebootstrapping - m
mammoth-pharmacist-54215
11/09/2015, 6:51 PMwe don’t really use docker much here at least not on dev laptops - h
happy-kitchen-89482
11/09/2015, 6:51 PMwe don’t use it at all yet - b
bored-art-40741
11/09/2015, 6:52 PMYeah, I was just trying to emphasize how hard it is in practice to truly clean all state - b
bored-art-40741
11/09/2015, 6:52 PMEspecially if you don't know the particulars of an org's configuration - b
bored-art-40741
11/09/2015, 6:53 PMSerious, somewhat tangentially related question: have we considered forking Ivy? - m
mammoth-pharmacist-54215
11/09/2015, 6:53 PMfair enough, it may make sense then to have this be a twitter internal plugin. @rough-minister-58256 this would be a candidate for the sentry work you were talking about. I’d like to be able to track how often people are actually using it. - r
rough-minister-58256
11/09/2015, 6:53 PMI think we need another thing for collecting large-format data/env dumps - r
rough-minister-58256
11/09/2015, 6:54 PMeventualyl would love to see a magic sploot dir, /sploot/support etc that you could just copy a tarball to and it would scribe that out to HDFS for offline analysis - m
mammoth-pharmacist-54215
11/09/2015, 6:54 PMthe general feeling at twitter seems to be that Ivy mostly works and isn’t a huge pain source right now for us. @bored-art-40741 did you have a reason you were thinking about forking? - r
rough-minister-58256
11/09/2015, 6:55 PM@bored-art-40741: I think the plan on our side is to eliminate Ivy by writing a generic multi-language resolver