Has SeaweedFS been considered as a distributed storage alternative to Ceph? From what I’ve read, it’s both faster and easier to manage.
It’s not really an alternative to Ceph, it’s an alternative to Ceph-FS and Ceph-Object.
This means that an Incus storage driver for SeaweedFS would not be able to hold actual instances, but just storage volumes and storage buckets.
That’s still valuable on its own, but the main benefit of Ceph is the RBD support which allows for instances themselves to be picked up from anywhere in the cluster.
Actually I am using SeaWeed with LXC Containers. As LXC are very efficient but LXC lacks a lot of functionality, which you need for cloud and cluster management I am looking for an alternative to LXC.
Can I use Incus for this setup somehow through a given Interface in SeaWeed Home · seaweedfs/seaweedfs Wiki · GitHub or does Incus need for this a driver like Ceph-FS? And if the answer is yes, how much effort in MT would it be to write one for SeaWeed?
I don’t have any experience with SeaweedFS at this point so can’t really comment on how it would natively fit in or how much work it would require.
But as far as Incus instances accessing it, it looks like you can connect to it through FUSE which would work fine from inside a container. Or alternatively you can use a disk device to share an existing mount with your containers.
Does Longhorn provide the equivalent block volume support alternative to Ceph?
Ceph is not without its issues and usually very slow. I was looking around for alternatives.
I just saw Longhorn supports iSCSI. Maybe lvmcluster can be used on top of it.
One idea discussed before is integration with Linstor, which is basically a manager for DRBD-over-LVM.
Although the management engine is in Java, the I/O path is a kernel module so is very fast; it’s basically RAID1 mirroring over the network. It requires DRBD9, but that does mean that each volume can be replicated up to 32 ways.
Yeah, I’m currently more interested in Linstor than seaweedfs, though I haven’t personally played with either of them directly.
It feels like the Linstor design is a closer fit to what we’d want for Incus.
Great to hear. Some alternative solution would be nice.
Btw, I did not know of Linstor before. I was reading on how it works and it does indeed look much better than Seaweed or Longhorn. And I understand now how it looks like a closer fit to incus.
Please ignore my Seaweed recommendation. I was looking for alternatives to ceph and it was the best one I had found so far. I now see Linstor looks much better.