Volumes replication & failover on 2 physical hosts

General
1

Hi,

I need to build a docker swarm with high-available persistent volume storage on 2-3 physical servers (8-16 GB Ram / 250GB SSD).

But after 2 days in the forums and articles, I did not find any good recommendation for the SDS (Software Defined Storage) part and integration with Docker Swarm.

Options I’ve seen so far:

  • Run NFS + DRBD + Scripts. But it does not seems efficient.
  • Run docker in Proxmox LXC. While Proxmox VE is an excellent product, I see low benefit in having LXC, especially since I’ll need to run docker inside LXC.
  • Run docker on top of Canonical OpenStack (MAAS, …): But seems overkilled.
  • Run ceph on the docker hosts directly (Ubuntu Server 18.04).
  • Run ceph/daemon with bluestore as a docker container on each docker host, and provide it a lvm partition.

Context:

  • I’m fine with command line (GUI is just a nice to have).
  • Ubuntu Server or Debian is preferred.
  • I want low overhead (so no full VM).
  • I’ll start with 2 servers, but can have a 3rd one in KVM VM if needed.
  • The restart of service must occur automatically if a host fails. (Swarm will take the container part, but I still need the storage part)
  • The solution must remain simple and easy to upgrade over time, and not require licenses.

The last solution (ceph inside docker container) seems the best. But did I miss something ?

1 Like
2

Hi, what did you end up doing? I’m in pretty much the exact position you were in back then.