A way to reconnect Discord after network failure?

[Wohlstand]

Hello!

It's a problem for a while that DiscordJS based bot might just stop the working even process itslef is still running and not failed. This happens always when there are network failures and disconnects. Is any way to handle the possible disconnect and run the reconnect from the scratch? Or at least self-kill process and let things like SystemD to restart the process. Recently I had two such accidents, like district-wide power surge that caused network fail, my server continued to work glad to the UPS thing. However, bot lost the connection and just hanged forever.

So, there are two questions:

• Any event that gets executed when connection with Discord was lost?
• Any kind of periodic methods like ping-pong to verify the connection with Discord and ensure the network is still accessible?

It's just to let bot self-restart or attempt to reconnect after network failures. Otherwise it's a pretty annoying case that bot is unable to self-repair in such cases.

[vakiliner]

The @discordjs/ws library automatically reconnects the bot when the network fails.

[Wohlstand]

If you can get us a reproduction sample / steps to reproduce, we can take a look. But discordjs/ws should always attempt to reconnect, no matter what

I forgot to mention, I ran all my bots with the --max-old-space-size=256 --expose-gc arguments of NodeJS, and the periodically ran function calls the garbage collector every half of hour. Is that possible the limited memory and garbage collector might cause such problem?

[Wohlstand]

The test case - the BastionBot that I mentioned before, just now I upgraded it to the latest stable DiscordJS and the code to use new API.

You just fill the JSON config file with the static guild and dedicated text channel, and the token to connect, bot is just replying to any "dance" word mentioned in messages sent at the dedicated guild channel. I ran the bot for 24/7. And the case I mentioned before, just let entire network to go off for a long time, like 5-10 minutes, and then try to see how bot will behave in the channels. And ye, before to start such test, just let bot work for a while before.

I do use this bot as a toy to let bridged users (who talk with Discord users from different messengers using bridge bot) verify does the bridge works or fails, so, they type "dance", and bot replies. If reply comes to them, the bridge works, but if no reply, it's possible failure of the bridge, or the test bot.

[vladfrangu]

Uhm. Aside of the fact you really shouldn't be limiting your processes to such small amounts of ram (and according to your github for bastionbot, you actually cap it at 128MB of ram) nor touch the GC (it can handle itself), the logs don't really show anything suspicious.

You should listen to debug events on your client and print those, and catch this error you're seeing (client.on('debug', message => {}) but you might want to filter out heartbeat messages), and ideally follow up with those here when you reproduce the issue

[Wohlstand]

I do such memory limiting knowing NodeJS is a greedy memory eater, and I don't want simple bots eat so much memory of the server (I would like to rebuild entire set of bots just into C++, but available SDKs for Discord aren't well developed).

You should listen to debug events on your client and print those, and catch this error you're seeing (client.on('debug', message => {}) but you might want to filter out heartbeat messages), and ideally follow up with those here when you reproduce the issue

I'll try to attach such listener and ocne bot will stuck with the networking again, I will have the evidence.

[Wohlstand]

And ye, I performed the test now, and seens it's too short time before it will go for a while, I'll set it up on my main server, and once it will hang, then I will dump a debug log.