02:31:57 <jelmd> Is there a way to obtain queue values `kstat unix::sysinfo`  without using kstat_* functions (in C)? I.e. getting the instant value instead of the "average"?
02:36:35 <jbk> i believe those represent the actual struct in the kernel that's updated by the kernel
02:36:56 <jbk> (I.e. i don't think it's processed before read by kstat)
02:39:06 <jelmd> kernel updates it every second via clock tick. thought on something like getloadavg().
19:44:46 <richlowe> jelmd: you're wanting to read sysinfo.runque directly?  You could do it via kvm if you were so inclined I guess, but I wouldn't.
19:45:10 <richlowe> also, `runque` is very fun to say, I don't care if it's meant to be two words
20:28:46 <jelmd> richlowe: yes, directly.
20:35:49 <richlowe> the only way I know of is libkvm which is really not great
20:36:05 <richlowe> I'm not immediately sure what avoiding the kstat is gaining you though.
20:43:12 <jelmd> Actually making the stats available via http (prometheus export format). If several clients query it (even if de-multiplexed) the average they get is not really the "same" if one tries to keep it simple, i.e. not tracking clients and just using the prev snapshot to calc the average.
20:53:29 <jelmd> Yes, libkvm looks a little bit too complicated for now.