App doesn't reconnect on lost connection

[pa-sowa]

When I suspend and later resume my computer it loses connection to the MQTT server and doesn't attempt to reconnect. I need to restart the app manually.
I see this error:
ERROR [client] Connect comms goroutine - error triggered read tcp 192.168.1.52:46248->192.168.1.45:1883: read: connection reset by peer

[W-Floyd]

I can reproduce the error message putting my computer to sleep, but it reconnects in some way. Perhaps I need to leave it asleep longer. Were there any more messages? How long did you leave it after resuming? There should be a 30-second keep-alive, perhaps we can try reducing that.

[pa-sowa]

It happens every time I put the computer to sleep for a longer time (ie. a few hours). It doesn't reconnect, neither within 30 seconds nor when I wait 30 minutes. There are no other messages in the log.

[W-Floyd]

Okay, thanks for the info. I'll see what I can do, started a new job last week so time is short.

[W-Floyd]

Sorry for not taking more of a look at this - I expect it might just be a matter of checking for a connection occasionally, and reconnecting if it goes down. If anyone wants to take a stab at it and make a PR, I'd be happy to see it.

Frankly, I haven't set this software up again since moving for my job, though I will in future. Until it bothers me, it'll probably keep being pushed down the line 😢

[dev-foo-bar]

May be a workaround. If you can implement an exit code we can use systemd to do the magic (simple restart).

[phallows]

Looking into this a bit more, I can see a couple of possibilities. It might be failing on "initial connection", which would be related to SetConnectRetry (default false). Or it might be failing because it is writing during a suspend, which would be related to SetWriteTimeout (default no timeout).
I'm going to deploy these settings to machines for a few days and see if it makes a noticeable difference.

[W-Floyd]

How did this go @phallows ?

[phallows]

It seems to be working quite well, but I've got an issue with an unrelated part of my process that makes it hard to verify. I think its probably worth a PR so that others can test, I'll submit that as soon as I can