Few questions on ops-manager backup process

  1. While backup ops manager creates a head db which is copy of the monogdb deployment I want to know the reason behind this. Why create another copy and increase space consumed for backup
  2. Ops manager takes snapshot of the head db which got created where Backup Daemon is running What is the type of snapshot?
  3. How ops manager maintains consistency in snapshots for whole cluster?

Whole process is explained here

Backup is created "elsewhere" (than originating data; of course) and that backup is kept "up-to-date". Time to time snapshots (moments in time) is created (and stored) from that data. In the "side" system keeps copy of opLog too, so it is possible to "roll-forward" from snapshot to specific moment in time (between snapshots).

  • Thanks for reply @JJussi but can you please be more specific your answer is very generic and which is the basic snapshot backup process. I need to know why do they create a standalone head db for every replica set and then they take a snapshot of that head db directories.instead of taking snapshot of original mongodb deployment
    – bash
    Jun 1 '17 at 5:05
  • Snapshot is "image" of situation in point of time. With consecutive snapshots it's delta between them. If you have replica set with multiple nodes, snapshot must be done at primary node. If primary changes, you cannot do delta to previous snapshot because it's on other physical node. If you collect all data (from replica set) to one location, you can do always delta.
    – JJussi
    Jun 1 '17 at 5:44
  • interesting !!!! This answers why they create a head db. Anything on other two questions ?
    – bash
    Jun 1 '17 at 6:36
  • 2. depends the snapshot type what you have select (blockstore, S3 or filesystem). 3. consistency is in opLog. We have all oplogs (from all replica sets) and every oplog entry have timestamp. opLog have only changes to data, as commands like create, drop, insert, delete, update...
    – JJussi
    Jun 1 '17 at 6:43
  • for 2.blockstore ,s3 and filesystem are snapshot store right? not a type and for 3. thats what I was thinking but snapshots are generated using head data directories ,so do they snap all directories one by one if so then how snapshot is consistent as oplog is being applied continuously
    – bash
    Jun 1 '17 at 6:54

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.