Data Replication Among Multiple IPFS-Clusters

Hello together,

I read a bit into the ipfs-cluster and have a few questions about it. Let’s say I have 12 nodes in a private network.

Is it possible to set up 4 clusters with 3 nodes each?
What happens if a node in a cluster has no free local memory and new data should be stored in the cluster?
If a node requests data from another cluster in which it is a member, is this data also distributed in the new cluster?

Thanks for the answer!

Sure, why not.

If disk space is gone, ipfs will error when pinning. If the IPFS repository uses more than the StorageMax configuration variable in the ipfs daemon, then things will just work as before, since hitting the StorageMax mark does not do anything in IPFS. When deciding where to allocate pins, cluster will give less priority to peers with less storage left. tl;dr: set StorageMax to something smaller than the available disk space in each peer to make sure you get even distribution of content in the cluster.

I’m not sure what “a node requests data from another cluster” means. A cluster peer pinning something that is pinned in another cluster? In that case, it depends on the ipfs daemons being configured correctly and connected among each others. If the 12 ipfs daemons are in the same ipfs private network then the content they are pinning will be accessible from any of the others, regardless of the cluster they are controlled with.

The cluster private network (formed by the cluster peers in a single cluster) and the ipfs [private] network are fully independent.

1 Like