| fungi | interestingly, nl01 hasn't logged stopping the rax-ord state machine provider for 877113 yet, not sure if that means it's not fully in use | 00:02 |
|---|---|---|
| fungi | we had quite a few spikes well above 10 building, leading me to suspect it is not | 00:03 |
| fungi | nope, my bad. i grepped wrong. it stopped at 18:48:11 (moments after deployment) | 00:04 |
| fungi | so the past ~6 hours have been with the new configuration, but we have a spike on the graph since then with 44 building nodes | 00:05 |
| fungi | also some others more sustained which are still well in excess of 25 | 00:06 |
| clarkb | fungi: note the old provider can continue running for many hours still. But the new provider should do most of the work. Previously there was no new provider until the old provider went away | 00:11 |
| clarkb | I'm going to remove gitea01-04 from haproxy by hand again in order to see how gitea09-14 cope | 00:12 |
| clarkb | and thats done | 00:14 |
| clarkb | there is a definite load average jump since taking the old servers out of the load balancer, but so far it seems reasonable | 00:27 |
| clarkb | if that changes you can easily erenable the old 01-04 bcakends by restarting haproxy (easy disruptive method) or running haproxy socket commands to reenable them (documented but 8 comments instead of 1) | 00:30 |
| *** dmsimard9 is now known as dmsimard | 17:58 | |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!