Al Chu

Assuming error is this (i ran by hand) ``` $ /usr/bin/loginctl enable-linger flux Error registering authentication agent: GDBus.Error:org.freedesktop.PolicyKit1.Error.Failed: Cannot determine user of subject (polkit-error-quark, 0) Could not enable linger: No...

some random thoughts, not necessarily for this prototype, but just thoughts for later (if this works out): - currently in `job-archive` we archive eventlog, jobspec, R, jobid, userid, ranks, and...

I hit a snag during my prototyping today and thought I'd do a thought dump before moving forward: - the major snag: the job-list/job-db module can be sent the journal...

> the major snag: the job-list/job-db module can be sent the journal event for "job inactive" (eventlog "clean" event) BEFORE the eventlog "clean" event is written to the KVS. So...

> What did the problem turn out to be with that? Mostly just hitting corner cases :-) It'd be easier if I could just grab the eventlog, vs having to...

> What if instead of trying to make sure the journal events are delayed until the associated kvs commit is complete, we added an entry to the journal for jobs...

> I didn't quite understand what you mean by via the job-list restart. I was presuming job-db would not need to scan the KVS on restart since it already has...

I finally got an experimental branch that completes tests in sharness (note completes, not passes :-). This prototype is garbage of course, tons of memleaks, tons of lazy cut &...

> If we load job-list in the rc1 script right after job-manager, then aren't we assured that it can capture all pending/running jobs from the journal? If not maybe we...

> 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...