Jim Garlick

Yeah I was just going to propose a PR that hides the subcommands, and also aborts if stdout is a tty (with a message to try flux-jobs(1)). Both John G...

Danger: you are about to inhale JSON! It was remarkable how patient Trent was about this. ("can you tell me how to convert the userids to names?")

A suggestion would be to leave off `guest.exec.eventlog`, stdio, annotations, and memos for this experiment. We can always add later. Also, a simplifying assumption may be that (unlike `job-list`) no...

I was going to suggest just reconstructing the event log from the journal. What did the problem turn out to be with that? On Mon, Apr 11, 2022, 4:30 PM...

> It'd be easier if I could just grab the eventlog, vs having to re-construct it ... and via the job-list restart. I didn't quite understand what you mean by...

> still need to scan the KVS on restart b/c job-list builds up its internal list of pending / running jobs from this vs. the journal. If we load job-list...

Ah gotcha. Right now all jobs are killed when an instance is restarted so we don't have that problem but we will. Good point. Let's see if we can work...

did we decide that job-archive could stick around as long as flux-accounting needs it (but ultimately flux-accounting would transition to this?)

Full chatty log attached. [picl0-crash-log.txt](https://github.com/flux-framework/flux-core/files/8135036/picl0-crash-log.txt)

Couple of good comments from @grondo on slack, captured before they scroll off into oblivion: - There is a TimeoutStopSec directive in systemd.service - "infinity" is a valid value -...