| sorbal | Hey everyone, is this the right place to ask a quick question about how octavia creates and manages amphora instances? | 21:29 |
|---|---|---|
| johnsom | sorbal: yes | 21:36 |
| sorbal | I noticed that while I am connected to an amphora instance via the horizon console I sometimes get disconnected. Then I noticed that it wasn't just a disconnect but the instance was being destroyed and recreated with different IP in the mngmt network. Is that normal behavior for the load balancer to recreate an amphora randomly? Or is there an issue? | 21:39 |
| sorbal | I am currently going through the logs to see if there is an error but for now I can't find something | 21:40 |
| johnsom | Check the health manager logs (controller process), there is some reason it thinks the load balancer is failed and it's attempting to repair it. | 21:46 |
| sorbal | thanks, I was looking at the worker log and although I could see the failover getting triggered I could not see the reason | 21:58 |
| sorbal | I have created my own image and I am running some tests that change the haproxy config and restart the haproxy service, somewhere in my process something goes wrong because I do get health message reports 0 listeners when 1 expected as warning in the health manager logs | 21:59 |
| johnsom | Yeah, health manager and the amphora agent monitor the haproxy stats socket to determine health. It will look for listeners (frontends in haproxy speak), etc. | 22:01 |
| sorbal | I know its a bad solution but since I am trying to develop something inside a custom amphora image is there a way to turn the health manager off? | 22:02 |
| johnsom | Yeah, just stop the process | 22:06 |
| johnsom | Statistics and status updates won't occur, but it also won't attempt to repair it | 22:07 |
| sorbal | Great! I paused the container (i am using kolla) and it worked. Thank you for your help johnsom | 22:32 |
| johnsom | Sure, NP | 22:32 |
Generated by irclog2html.py 2.17.3 by Marius Gedminas - find it at https://mg.pov.lt/irclog2html/!