You are viewing a plain text version of this content. The canonical link for it is here.
Posted to oak-issues@jackrabbit.apache.org by "Amit Jain (JIRA)" <ji...@apache.org> on 2014/04/03 14:29:14 UTC

[jira] [Commented] (OAK-1500) Verify restore to revision on MongoNS

    [ https://issues.apache.org/jira/browse/OAK-1500?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13958761#comment-13958761 ] 

Amit Jain commented on OAK-1500:
--------------------------------

Tested the following scenarios by following the backup restore for mongoo db using [mongodump/mongorestore utility | http://docs.mongodb.org/manual/tutorial/backup-small-sharded-cluster-with-mongodump/]. This was tested with a small 2 shards setup, by following the [mongo shard setup | http://docs.mongodb.org/manual/tutorial/deploy-shard-cluster/]. But ids being provided by the test redirected the writes to only one shard.

* Backup after a node being shutdown and restored.
* Large writes happening and randomly taking a backup, and then restoring the db.

With the _lastRev recovery in place the start of the node after the backup looked good and in a consistent state. With reads/writes succeeding and _lastRev of root getting updated to the revision of the latest node written.

> Verify restore to revision on MongoNS
> -------------------------------------
>
>                 Key: OAK-1500
>                 URL: https://issues.apache.org/jira/browse/OAK-1500
>             Project: Jackrabbit Oak
>          Issue Type: Improvement
>          Components: mongomk
>            Reporter: Alex Parvulescu
>            Assignee: Chetan Mehrotra
>            Priority: Critical
>             Fix For: 0.20
>
>
> Following the discussion on OAK-1339, the backup will be controlled from Mongo directly, but we still need to verify 2 things:
>  - ongoing transactions will be discarded on restore, the head has to point to the latest stable revision
>  - as an added benefit, the restore could happen to an older revision (see the sharded setup where a node can get ahead of the others between the moment the backup starts and when it will finish across the board)



--
This message was sent by Atlassian JIRA
(v6.2#6252)