Simplest way to pool blocks and datastore among disks?

Help Coding
1

Sure zpool or glusterfs could do the job, but I’m trying to avoid overhead from more IO layers.

It looks like this should be possible by adding entries to ipfs config Datastore.Spec.mounts, but I’m having trouble with the syntax. Should I copy the child block and change the path, or copy the whole first mount entry and change the mountpoint?

[
  {
    "child": {
      "path": "blocks",
      "shardFunc": "/repo/flatfs/shard/v1/next-to-last/2",
      "sync": true,
      "type": "flatfs"
    },
    "mountpoint": "/blocks",
    "prefix": "flatfs.datastore",
    "type": "measure"
  },
  {
    "child": {
      "compression": "none",
      "path": "datastore",
      "type": "levelds"
    },
    "mountpoint": "/",
    "prefix": "leveldb.datastore",
    "type": "measure"
  }
]

Ideally I could let IPFS fill entire dedicated partitions, and possibly some sort of automated load balancing across mounts.

By the way, EDITOR=nano ipfs config edit hangs on exit even without saving changes…

1 Like
2

I’m handling all this through LVM in my development environment. I believe there are some issues with IPFS and multi-datastores however I’m not 100% certain on that so that is why im using LVM.

1 Like
3

Thanks for the suggestion! I’m currently evaluating bcachefs as a modern general-purpose filesystem which could also solve this nicely.

4

interesting, i’ll take a look at that as well. Thanks.

5

At the moment, there are no datastore adapters for sharding (just mounting datastore prefixes).