bug/regression: node ca be started on multiple clusters

[fbarbu15]

A node can be started now with multiple clusters by using ex: --pubsub-topic=/waku/2/rs/2/0 --pubsub-topic=/waku/2/rs/3/0

docker run -i -t -p 10255:10255 -p 10256:10256 -p 10257:10257 -p 10258:10258 -p 10259:10259 harbor.status.im/wakuorg/nwaku:latest --listen-address=0.0.0.0 --rest=true --rest-admin=true --websocket-support=true --log-level=TRACE --rest-relay-cache-capacity=100 --websocket-port=10257 --rest-port=10255 --tcp-port=10256 --discv5-udp-port=10258 --rest-address=0.0.0.0 --nat=extip:172.18.222.250 --peer-exchange=true --discv5-discovery=true --cluster-id=2 --metrics-server=true --metrics-server-address=0.0.0.0 --metrics-server-port=10259 --metrics-logging=true --relay=true --nodekey=30348dd51465150e04a5d9d932c72864c8967f806cce60b5d26afeca1e77eb68 --pubsub-topic=/waku/2/rs/2/0 --pubsub-topic=/waku/2/rs/3/0

And you can relay and receive messages on both clusters

This was not possible on v0.26.0

[gabrielmer]

Seems that this happens because if we don't define a cluster-id, it implies that we use cluster-id=0 and we override its pubsub topics for "/waku/2/default-waku/proto"

nwaku/waku/factory/waku.nim

Lines 109 to 112 in 840e012

	# cluster-id=0
	of 0:
	let clusterZeroConf = ClusterConf.ClusterZeroConf()
	confCopy.pubsubTopics = clusterZeroConf.pubsubTopics

If we set a cluster-id in which we don't override the pubsub topics, for example running

./build/wakunode2 --pubsub-topic=/waku/2/rs/2/0 --pubsub-topic=/waku/2/rs/3/0 --cluster-id=2

we do get the error

ERR 2024-05-14 15:35:59.034+02:00 Waku initialization failed                 topics="wakunode main" tid=10401565 file=wakunode2.nim:79 error="Failed setting up node: failed to create record: failed to parse pubsub topic: use shards with the same cluster Id."

@fbarbu15 wdyt?

[fbarbu15]

Ok, makes sense. Closing as not a bug then. Thank you

[alrevuelta]

Reopening. Doesn't matter how the node is configured, a node should never be running in multiple clusters. In the case @gabrielmer states its protected, but in the one you showed its not.

[gabrielmer]

Reopening. Doesn't matter how the node is configured, a node should never be running in multiple clusters. In the case @gabrielmer states its protected, but in the one you showed its not.

Not sure I understand. If we start the node in clusterId 0, the pubsub topics are going to be overriden for the default one and the end result is the node running in clusterId 0 with the default pubsub topic. The protection is still in place, but because the pubsub topics get overriden for a valid one then there's no error.

Should we not override the configurations for clusterId 0?

[alrevuelta]

After discussing with @gabrielmer the behavior is expected. Cluster-id=0 overrides whatever topics are configured. We could warn but that's expected.

We can close since named sharding will be revomed and this won't happen afterward.
#2163