ping @jeremyeder @jfrazelle @philips @tianon

what is the default, just curious, and i will try this out

@lsm5 I would love to see something explaining why Infinity in your commit, just so we remember. I have no objections though.

Indeed, I think this PR is missing pretty much 100% of the rationale. 👍

Ping @lsm5 can you please clarify your rationale?

so the default core size on fedora/rhel systems is 0 and current rpm packages weren't generating any core dumps. Got this request via our support folks, and they'd like as much detail as possible in a core dump.

Sorry about the rushed up PR, but that's all I have for now :| , I'll try and grab more info though. Perhaps @jeremyeder has more insight into this.

Yes, this came from our support folks, but it benefits everyone. We're aware of the shortcomings of go+gdb, but still, there is some information that can be gleaned from a core file, which is a lot better than logs, debug logs or posthumously adding printk's and trying to reproduce.

So we'd like to suggest the docker unitfile be configured to write a core so we can perform root cause analysis in case any crashes occur in the field. "infinity" is about ensuring we get a complete core.

Ok, thanks; seems reasonable to me! 👍

LGTM

ping @philips

guess we also need a GOTRACEBACK=crash to enable core dumps, but that's probably set somewhere else, set via /etc/sysconfig/docker for fedora and its downstreams

also, I'll send a follow-up commit here with info from @jeremyeder in the commit message

lgtm.

The side-effect is that all docker containers get this limit internally too, right?

Thanks to everyone for looking at this.

Yes, @philips that's correct...softlimit=0 is overridden because RHEL has a core-catcher pipe-command called abrt. We're taking care of disk space hogs there, plus a bunch of other tweaks.