The backup is node independent so when restoring an entire 3 node cluster, you can restore the backup on each. What you don't want to do is only restore on one node with the other 2 with empty DBs and bring up all 3 at the same time. In that case, the 2 with no data may form a majority and the cluster comes up with no data. You can stagger them, ie restore on 1, bring 2 up and verify they're synced, then bring up the 3rd. But it's easiest to just restore on all 3.
On Thu, Jan 12, 2017 at 11:03 AM, Miguel Angel Muñoz Gonzalez < [email protected]> wrote: > Great to know this info, thanks everyone! > > > > I am afraid it is not possible to restore backup while ODL is running, > correct? > > > > Just one more thing to confirm: In a cluster, it would be enough by > backing up one node, while restoring should be done in every node in the > cluster, am I right? > > > > Thank you very much, > > Miguel Ángel. > > > > *From:* Tom Pantelis [mailto:[email protected]] > *Sent:* jueves, 12 de enero de 2017 15:08 > *To:* Lori Jakab <[email protected]> > *Cc:* David Suarez Fuentes <[email protected]>; Miguel > Angel Muñoz Gonzalez <[email protected]>; > [email protected]; [email protected] > *Subject:* Re: [controller-dev] Backup and Restore > > > > yes - directory. Backup file name doesn't matter. You put the file in that > directory and it's read on startup and then deleted. However the journal > and snapshots must be empty or else it won't restore. > > > > On Thu, Jan 12, 2017 at 8:40 AM, Lori Jakab <[email protected]> > wrote: > > On Thu, Jan 12, 2017 at 3:01 PM, Tom Pantelis <[email protected]> > wrote: > > You restore by putting the backup file in > $KARAF_HOME/clustered-datastore-restore > (create it) > > > > I assume `clustered-datastore-restore` is a directory. Does the backup > file name matter? > > > > -Lori > > > > , delete journal and snapshots directories and start karaf. > > > > On Thu, Jan 12, 2017 at 5:20 AM, David Suarez Fuentes < > [email protected]> wrote: > > Hi, > > > > Thanks Lori to point out to this API method. By looking at the methods > offered by the cluster-admin module, what I cannot see is the "restore" > method. Can someone tell us where we can find this? > > > > Thanks in advance, > > David. > ------------------------------ > > *De:* [email protected] < > [email protected]> en nombre de Lori Jakab < > [email protected]> > *Enviado:* miércoles, 11 de enero de 2017 21:31:44 > *Para:* Miguel Angel Muñoz Gonzalez > *Cc:* [email protected]; controller-dev@lists. > opendaylight.org > *Asunto:* Re: [controller-dev] Backup and Restore > > > > On Wed, Jan 11, 2017 at 8:39 PM, Miguel Angel Muñoz Gonzalez < > [email protected]> wrote: > > Hi everyone, > > We are trying to implement a backup/restore mechanism for ODL. I have not > found much information about it in pipermail history except some questions > and brief proposals (such as generating an snapshot and storing in a > particular file, using an external database,…). I would appreciate if > someone familiar with the topic could give us some hints on it: > > > > - Is it possible to backup MDSAL Configuration datastore as of > today? (I suppose it’s a matter of backing up leveldb files) > > > > - If so, how can we guarantee that the backup is correct and/or > consistent while it is running traffic and datastore is being modified? > > > > - Supposedly there is a working mechanism consisting on: > stopping ODL, copying the files manually and start it up… However, it would > nice to know if there a more sophisticated mechanism to backup the > datastore, especially if it does not imply stopping ODL. E.g. a particular > API or tool that can be called to perform this activity? > > > > I hope more knowledgeable people will give you a detailed answer, but > there is one RPC that allows creating a backup of ODL while it is running: > > > > POST /restconf/operations/cluster-admin:backup-datastore > > > > Where the message body looks like this: > > > > { > > "input": { > > "file-path": "/tmp/datastore_backup" > > } > > } > > > > Since the datastore is transaction based, I assume this will save data up > to the last successful transaction, but don't quote me on that. > > > > One of the main issues with backup/restore, AFAIK is the possibility that > a schema (YANG model) may change between the backup and restore, and how to > handle that. > > > > -Lori > > > > _______________________________________________ > controller-dev mailing list > [email protected] > https://lists.opendaylight.org/mailman/listinfo/controller-dev > > > > > > >
_______________________________________________ Discuss mailing list [email protected] https://lists.opendaylight.org/mailman/listinfo/discuss
