Guest User

Kurt - Migrating to ZFS

a guest
Nov 15th, 2016
Not a member of Pastebin yet? Sign Up, it unlocks many cool features!
  1. Hi Allan and Kris,
  3. This question is probably mostly directed at Allan... thank you both for the show, which I listen to pretty much every week. Here is my situation. I'm a mostly new user of ZFS, and I've pretty well mastered the topics in the first book by Allan and Michael. I have a collection of hard disks that contain non-ZFS clones of my user data from various time points. I'd like to turn those into ZFS snapshots so that I can get rid of this collection of boat anchor disks in a box, while still preserving the ability to reach back in time to pull out an old file if necessary. I had originally thought that I would bring them into ZFS as separate datasets with deduplication turned on, and even started down that path, but quickly found that this was overwhelming my computer's RAM and ability to be responsive (even though the deduplication factor was pretty high). I learned why "Friends don't let friends de-dupe."
  5. So, I'm thinking of another way to accomplish this, and thought I would run it by you for feasibility. I thought that one way to accomplish this would be to bring the oldest clone into ZFS as a dataset. Then create a snapshot of that dataset. Then rsync the next oldest clone with the dataset. Then take another snapshot. I would continue forward in this way, doing rsync and taking snapshots until the final rsync would be for my current user data. At that point I would have a series of snapshots going back in time that represent each of the hard disks. Since the user data doesn't change much from slice to slice in time, this shouldn't take up much space, right? And since ZFS doesn't have to manage a dedupe table, it shouldn't eat all my RAM, right? The idea being, if I want to pull a file from an old snapshot I can just go into the hidden directory for that ZFS snapshot and grab the file; if I want to re-create an entire slice (because I perhaps want to access a large number of files) I could do a zfs send operation to load that slice into a new dataset? I'm not entirely schooled on the possibilities of send yet (as I haven't bought the 2nd Jude/Lucas book on ZFS). I also like the fact that these old snapshots are read-only, and thus not subject to accidental erasure or change.
  7. Thank you for any advice you have, or pitfalls you might want to point out.
  9. Kurt
RAW Paste Data Copied