3

My goal is to find a distributed filesystem on Linux that supports ZFS-like lightweight snapshots and snapshot clones. This StackOverflow question expresses what I'm looking for pretty well. I'm trying to figure out whether Lustre + ZFS is the solution.

If I set up a ZFS-backed distributed Lustre filesystem, will I be able to snapshot it, and then mount and clone those snapshots? Or would the ZFS snapshot features operate at the level of individual OSS's rather than across the entire distributed filesystem?

My research so far has steered me toward Lustre + ZFS as a promising option, but the consensus on the internet seems to have been that ZFS on Linux was too beta until the very recent 0.6.1 release. The current Lustre docs do briefly mention ZFS as an optional backend to Lustre. Possibly as a result of the very recent maturity of ZFS on Linux, I haven't found any reports of Linux deployments of Lustre + ZFS except LLNL's Sequoia, and I haven't managed to find documentation related to Sequoia that answers my question.

Anand
  • 31
  • 2
  • You've already answered your own question - no. I do have a question, though; what's your use-case, that you're after a distributed filesystem on Linux with CoW-like cloning and snapshots? – Nex7 Apr 04 '13 at 23:54
  • It's an HPC setup. Programs are often 'one-offs' that don't get packaged and versioned, but nevertheless live and evolve for a long time. It's nice to be able to roll back to points in the past where they were known to produce certain results. – Anand Apr 05 '13 at 16:09
  • Andreas Digler answered my question here: https://lists.01.org/pipermail/hpdd-discuss/2013-April/000126.html. Looks like the answer is no, I couldn't do filesystem-wide snapshots. – Anand Apr 04 '13 at 23:03
  • It is worthwhile to amend this answer to note that Lustre 2.10+ has utilities to create and mount ZFS snapshots as Lustre filesystems, namely `lctl snapshot_create`, `lctl snapshot_mount`, etc. This is intended for use by e.g. periodic snapshot scripts via `cron` or similar, and then mounted (possibly via `automount`) on login nodes for interactive use (e.g. recovery of deleted files). Since the snapshots prevent space from deleted files from being released, the number of snapshots that a site keeps, and how long they are kept, depends heavily on the amount of free space available in the files – LustreOne Apr 05 '18 at 19:58

0 Answers0