1

We have a Windows 2012 R2 file server on a EC2 instance, with DFS network shares. The overall data takes about 600-800GB.

I would like to somehow keep a copy at 15 minutes intervals for the lastest 72 hours, to cover the use case of a user asking urgently to recover a file deleted or corrupted accidentally.

Is there any solution for AWS covering this particular use case of users screwing up with files? For disaster recovery we are using daily AWS snapshots and multisite DFS replication, but we are looking for something to recover a very recent file with a very short turn-around.

  • Maybe Amazon S3? Maybe something like Crashplan with versioning? There are probably innumerable ways to achieve this. – joeqwerty Jun 19 '16 at 16:09
0

The most simple would be to just snapshot the EBS volume at 15 minute intervals, and expire them after 72 hours. To restore, you'd just create an EBS volume from whatever snap you need, attach it to one of your servers, mount it, then recover the file.

All of this is trivial to automate, and will provide quick local recovery of files.

| improve this answer | |
  • I was under the impression that snapshotting the whole volume was somewhat taxing in terms of EBS performance (i.e. I though snapshotitng should be done just once or twice a day iduring off-peak times). Is the theory of snapshotting being a resource hog a urban legend? – Pep Jun 19 '16 at 17:56
  • Depends on how snapshots are implemented. With EBS, the first snap will take a while, and may affect performance, but all subsequent snaps are incremental, and will be very fast, especially with 15 minute intervals. – EEAA Jun 19 '16 at 18:01
  • Unless you remove the mounted volume first (or stop the instance in the case of the boot volume) though, you are not guaranteed to get an uncorrupted version of a file on a snapshot. Files continually open and frequently written to are more likely to be corrupted. – Karen B Jun 19 '16 at 18:21
0

I have a few ideas, none of them are probably going to be quite what you want, but it may trigger better ideas in others

  • BitTorrent Sync. Sync the server to another computer, with it set to upload only, not delete.
  • Dropbox, which can keep a version history
  • Attic, a de-duplicating backup program, but running it every 15 minutes on that much data might mean it never finishes.
  • CrashPlan, though every 15 minutes would be much more than it defaults to, and it may not finish before the next run starts.
  • Perhaps some kind of sync to S3

With most of these you'd have to work out how to remove files that aren't needed after your 72 hour window.

If you change your requirement out to hourly snapshots some of these become more practical.

Ideally you want something supported by AWS. Nothing comes to mind right now. EBS volumes are safe to use while snapshots are being made, but you'll experience degraded performance.

| improve this answer | |
  • We're currently doing it at 15 minute intervals on premises with VSS and DPM, but I wanted to see if there was something more affordable in terms of requirements and license costs. – Pep Jun 19 '16 at 22:46

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.