Hmm, there's something wrong with the new [[agora federation]] service 馃槉 it was using 16G of RAM after running for only about eight hours. I need to understand the leak and/or limit its maximum runtime in systemd :)
Discussion
Loading...
Hmm, load average was 17 in [[thecla]] and then I lost connection. Then I got back in and load average was 45.
Now everything is smooth and fast again. Hmm.
Maybe cache misses/cold loads have become very costly for some reason and restarts are bringing it down? Need to analyze logs.
O-oh, now everything froze and there was no obvious traffic/utilization spike before it froze (as shown by htop I was watching)
Oh, I bet it's OOMs. On closer inspection memory is tight and swap is full.
Hmm, there's something wrong with the new [[agora federation]] service 馃槉 it was using 16G of RAM after running for only about eight hours. I need to understand the leak and/or limit its maximum runtime in systemd :)
Fast #Agora is now back, serving random nodes within ~1s on average 馃槉