hhmx.de

Menel

Menel
@menel@snikket.de

Föderation · Fr 21.02.2025 13:04:27

/ activitypub noob :

job fifo size (cur): 655
job fifo size (peak): 1291
thread  state: waiting
thread  state: output
thread  state: output
thread  state: output
This number is decreasing over some minutes after I made a post,
I assume it is my instance delivering this to all subscribers, is that correct?
And the last thread will stay idle to maybe wait for incoming requests, so that they can be answered too?
I've never seen another status then waiting or output so far.

I guess if I post a picture that might then happen and all workers will be busy? Maybe even too busy to keep up?

Is that assumption about how the waiting worker and what it is for correct?

If yes I guess I'll increase the threads, if they are intentional low for low ram systems, that's not my issue with snac. (I've not much ram, but I assume snac is by default tuned to be very very very conservative?)

(adding Pic to simultaneously test my theory)



Two Monsters made out of wood and grass, standing on long legs in a lake.

(Medien: 1)

0x   1   1x

The Real Grunfink

The Real Grunfink
@grunfink@comam.es

Föderation · Sa 22.02.2025 07:50:03

Hi. You're right: the first fifo size number (cur[rent]) will be counting down until all outgoing messages are sent. The peak is just gossip about how long this queue has been.

Basically, all threads work in the same way. There are four states: stopped, waiting, input or output. It's almost impossible to see the first one, as it's only set while initializing or shutting down; input is also seen rarely, unless the instance posting the data is specially slow. So they are usually waiting except while in the «send storm», when they are mostly in output mode. Anyway, even in this case, an incoming message has preference over everything, and the first available thread will serve it.

sets up as many threads as cores are available, but being that the thread work is basically network I/O bound (there is very few CPU work to be done), you can increase the number of threads to twice or more the number of cores and performance will be slightly improved.

Interesting photo!

0x   0   0x