I’ve read that Raft requires its peers to have low network latency for replication. If I have several IPFS nodes across different locations, can I join them in a single IPFS cluster? Or is it advisable for nodes running in separate regions to have their own ipfs cluster?
ipfs-cluster allows you to configure a bunch of raft options so that it can run in high latency environments (it increases the failure thresholds basically).
They are in the
raft section of the configuration, but the defaults are very conservative already and it probably works with them. We have been running cluster with 10 peers around the world without issues.
I have recently started rewriting the docs so I’m going to point you to some documentation drafts:
- Config file reference https://cluster.ipfs.io/documentation/configuration/#the-service-json-configuration-file
- Config file tweaks (with advice for high latency scenarios): https://cluster.ipfs.io/documentation/deployment/#service-json-configuration-tweaks
Thanks for the explanation @hector. The configuration docs you have sent is perfect and it is what I’m looking for.