I'll drop the param for now, that is something I can workaround on my end. The bad path for the versionlock list is problematic in a lot of ways

PR updated

Yeah, I agree on the path being a problem (because it's broken)

@Pat Riehecky Looks like you have the same issue in

post_transaction_action

😕

That came with the existing codebase, but I'll get it fixed up and re-pushed shortly

Well, that's neat 😐

pushed

Looks good to me, @vchepkov?

Merged

ewoud: Do you think that

daemon-reload

should simply be implied no matter what? The only reason I added the parameter was to preseve the current (in my opinion broken) state if someone wanted it for whatever reason

tvaughan: I think it's a good question - ideally speaking you manage the service with Puppet and then Puppet 6.1.0+ should take care of the daemon-reload is needed - that said, we did find out that cases where you ensure something is absent, that won't work

also, having a drop in file but not managing the service would break, but I'm not sure why you'd do that (the manual restart in service_limits is still weird and something I'd deprecate)

At this point, I know that it works with GDM and NFS (the hard edge cases) so I'd like to get it through and then we can figure out if puppet is broken or something else is going on

so I do wonder in which other cases it would be needed

Both GDM and NFS actually have other auto-spawned items that need to happen in between the file drop and the service

GDM and NFS on which OSes (and versions)?

So, there's systemd magic going on as well

EL7 and 8 (and I suspect 9)

GDM because we're working around

hidepid=2

issues

Basically, there are more situations that need the intermediate reload than you might think at first glance.

I'm questioning if the default should be true or false - false would mean you don't apply a workaround so technically a bit faster but true catches more issues

That said, we could disable it by default and have users call it explicitly. That would reduce the execs but leave you scratching your head as to why things aren't working the way you expect

(and/or puppet itself isn't doing what it's supposed to)

ewoud: The daemon reload execs are fast and (unless you have a TON of dropins) you're really not going to notice

can I just rant that I hate that systemd needs daemon-reload at all? …

100%

I'd be ok with defaulting to true for now, but can we open an issue with the notes we have here so we at least have it written down? I don't count IRC/Slack as persistent

I mean, I guess I get it. If you're modifying a bunch of stuff, it doesn't want to go making a mess

Sure, I can copy/pasta it into a new ticket

thanks