mirror Of cluster in different region

hoseiney

we need to have complete mirror of vertica cluster to different region(other province)
does vertica handle this scenario ?

Jim_Knicely

Sure! Check out "Replicating Tables and Schemas to an Alternate Database" here:

https://my.vertica.com/docs/9.0.x/HTML/index.htm#Authoring/AdministratorsGuide/BackupRestore/ReplicatingtoanAlternateCluster.htm

sKwa

Hi!

Do you need mirroring or replication? Its not same thing:

mirror != replica

Jim_Knicely

Yeah, to @sKwa's point, with replication there'd be a delay on when the remote cluster is updated.

hoseiney

@sKwa @Jim_Knicely
we want to have a second copy of our data in different place that contains latest changes
we do not want immediately switch between servers

Jim_Knicely

If there is no issue with having "some" missing updates in cluster 2 (i.e. when cluster 1 fails prior to changes being applied to cluster 2), then replication seems like an acceptable solution. It's a built-in tool and you won't have any additional third party tool costs.

There is also a copy cluster feature, but here cluster 2 could fall way behind cluster 1 with regard to data modification to cluster 1 depending on how often you actually copy the cluster...

Or, if there is an issue with having "some" missing updates in cluster 2, you could apply all of your DML (Updates, Deletes and Inserts) to both clusters at the same time. The problem there is you'll have to modify your code base to do that. Also, you'll need to consult your sales rep as in this case both clusters will most likely count against your Vertica license.

Check out this blog post which reviews your Vertica built-in options:

https://my.vertica.com/kb/Vertica-Backup-and-Restore-Whats-Your-Use-Case/Content/BestPractices/Vertica-Backup-and-Restore-Whats-Your-Use-Case.htm

sKwa

Hi!

does vertica handle this scenario ?

Out of a box - no.

contains latest changes

1. You can achieve it with proxy, i.e. your DML applied on both servers simultaneously.
2. Or write your own tool that will observe for changes on server A and applies these changes to server B.

But proxy is a better solution.

PS: vbr.py(Vertica backup/restore tool) based on rsync tool and its not recommended to use it for replica.

Jim_Knicely

I totally agree that "DML applied on both servers simultaneously" is the best option, which is why I pointed that out in my earlier response

But, replication is incremental. The first replication of an object replicates the entire object. Subsequent replications copy only data that has changed since the last replication. So the replicate will have the latest changes after the replication process completes.

We have many clients that are using replication successfully. Although it may not be the best option for "mirroring/copying" an entire database... It's best use case is for replicating key objects.

sKwa

@Jim_Knicely

We have many clients that are using replication successfully.

Your opinion is biased. It would be strange to hear from you the opposite opinion.