Conversation
Notices
-
Recently discovered that someone I know clobbered a package installed binary, and when I asked why we didn't just build an rpm I was told building rpms is complicated and time consuming. This is after having the clobbered binary disappear because of package updates and there being almost no indication. Personally I don't think rpms are particularly difficult to build. I do think weird cluster failures that don't have an immediate explanation are time consuming though.
-
I'm not even sure how many times this bit us. Total time spent investigating bullshit is probably higher than what would have been spent modifying the rpm. I tried pleading we should just do it and was told we won't have the system long enough for it to be worthwhile. I'm going to recommend we at least make the binary immutable at the filesystem level so if someone doesn't follow the convoluted workaround update procedure they can't install the pending update.
-
@white_male i don't know if I could. It's been in place for a few years at this point . It probably hasn't been a giant sink, but it's a system that when it is down (and won't fail over) becomes a single point of failure for hundreds of sites. Much lower tolerance for bullshit on those pain points in the network IMO.
-
@thatguyoverthere Write down detailed time lines of the bullshit in downtime and man hours.
-
@white_male a bit concerned about stirring up more work at the current stage. I think since the system is on an eol schedule I'd be happy making the binary immutable at this point just to create pain for anyone that can't follow the upgrade procedure we now have documented. At least it won't prevent critical services from starting up.
-
@thatguyoverthere You definitely can starting now + whatever memories you have intact. I don't know how much you care about possibly shaking things up for yourself.