Failed to ping server fix

[msaizar]

The server will still be in a running state after ping failure, and the autoscaler will keep trying to reprovision it. The stopped time is set, but the state is not updated. This puts it in a stopped state, similar to https://github.com/drone/autoscaler/blob/master/engine/collect.go#L108-L109 and https://github.com/drone/autoscaler/blob/master/engine/reaper.go#L121-L122

[bradrydzewski]

good catch. I posted a minor comment with regards to which status I think we should use, but otherwise should be good to merge.

[msaizar]

Updated. The reaper took care of it eventually I think but I'll look into the GC, I haven't tried it yet, thank you! I also noticed that the pinger is not closing connections. On long running agents, you can see a lot of docker connections from the autoscaler with netstat -an, one for every 10 minutes the agent was up. Haven't figured out a fix yet but I'll open a discourse topic on that

[bradrydzewski]

thanks!

The reaper took care of it eventually I think but I'll look into the GC

yes, sorry, I was actually referring to the reaper

I also noticed that the pinger is not closing connections

the Ping function is actually the Docker Client code. So if there is a connection leak it would unfortunately be in the Docker client library. So we would need to send them a patch.