The document shows an example of recovery from an OODP snapshot in the case of a total zpool failure. The data is recovered to a newly created pool.

The loss of a zpool is simulated by deleting it completely for the purpose of this document only. Do not remove the zpool on the production server.

Configuration

Two Open-E JovianDSS instances are configured:

• node-prod with the production Pool-prod, a dataset on the zpool and a share (samba in this case)

• node-oodp as OODP snapshot target with Pool-oodp and a dataset

node-prod has a production backup task configured:

Snapshots are made on production node node-prod and replicated to node-oodp:

 Pool-oodp contains the dataset dataset-oodp which is marked with “Ds” (Destination):

 We are simulating the zpool loss by deleting it (Just for the purpose of this article. Do not remove the production zpool!).
After that we create a new zpool that will replace the old one after the simulated major failure:

There is no backup task on node-prod:

Dataset on node-oodp is still marked as “Ds".
To use this dataset as the source for oodp it is necessary to remove the Destination source path denoted by the "Ds" symbol:

And create a backup task on node-oodp with the destination server node-prod:

     After adding a backup task dataset-oodp is set as A B labeling it as a snapshot source:   

    After the first snapshot cycle (1 minute in this case) snapshots from node-oodp are replicated to node-prod (with the newly created Pool-prod-new):

When snapshots are replicated, remove the backup task on node-oodp and check if the dataset is not marked as "Ds":

Now it is possible to rollback the snapshot on node-prod - this way the data from the snapshot is recovered on the production node. The data can be accessed by adding a share to the recovered dataset:

Connect to the share and check if the data is available.

At this stage Pool-prod is recovered as Pool-prod-new.
It is possible to recreate backup tasks to node-oodp as usual for backup.