This quarter is going to be focused on bringing ipfs-cluster to life as a usable product in the IPFS ecosystem. That means:
- It shouldn't hard to crash
- It shouldn't lose track of content
- It should play well with go-ipfs
- It should support a replication-factor
- It should be very well tested
- It should be very easy to setup and manage
- It should have stable APIs
On these lines, there are several endeavours which stand out for themselves and are officially part of the general IPFS Roadmaps:
- Dynamically add and remove cluster peers in an easy fashion (ipfs/team-mgmt#353)
This involves easily adding a peer (or removing) from a running cluster. ipfs-cluster-service peer add <maddress> should work and should update the peer set of all components of all peers, along with their configurations.
- Replication-factor-based pinning strategy (ipfs/team-mgmt#353)
This involves being able to pin an item in, say, 2 nodes only. Cluster should re-pin whenever an item is found to be underpinned, which means that monitoring of pinsets must exist and be automated.
- Tests (ipfs/team-mgmt#360)
In the context of the Interplanetary Test Lab, there should be tests end to end tests in which cluster is tested, benchmarked along with IPFS.