Is there any already implemented way to shard in twilight 0.17 or is rolling your own sharding manager the way to go?
#Sharding in a prod bot
pallid saddle
opaque topaz
Is there any already implemented way to shard in twilight 0.17 or is rolling you...
The example is not up to date ?
https://github.com/twilight-rs/twilight/blob/main/examples/gateway-reshard.rs#L29
rapid dragon
that example is for something else, there is no difference how you should do sharding between 0.16 and 0.17
pallid saddle
that example is for something else, there is no difference how you should do sha...
Yeah, I was just asking in general, not spesifically 0.17 since I rolled my own manager when using 0.16 but I’m asking what’s the idiomatic way to do sharding just in twilight in general
rapid dragon
I just run each shard seperately with a small bit of shared state, I have not really had a need for a manager.
pallid saddle
I just run each shard seperately with a small bit of shared state, I have not re...
What about shard crashes/disconnects?
rapid dragon
Disconnects are handled internally in the shard, and generally crashes should be avoided
dapper elm
You can observe tokio task panics with JoinSet and track tasks through task::Id
Each shard should be in their own task for optimal performance fyi. How was your shardmanager setup? Channels?
pallid saddle
Each shard should be in their own task for optimal performance fyi. How was your...
No, I spawned tasks without tracking them in joinset when doing 0.16, I'm doing a small rewrite of my shard manager for a new bot with 0.17 now though
But a JoinSet is probably better in this case (and in all cases)
dapper elm
Have you run into shard crashes? The shard task should be very small, essentially just a next-spawn loop. I mean there's not much that can panic there
Otherwise I'd use mpsc channels for state management. It fits in really nice since the shard is cancellation safe
pallid saddle
Have you run into shard crashes? The shard task should be very small, essentiall...
yeah I did once surprisingly, and it for some reason could not recover with next_event, I don't have the logs but this was a stress test of over 100 shards so maybe not unexpected