How should we prepare XDCR for destination IP failure?

Couchbase Server
1

Hi All,
When we set XDCR, we set the destination to IP of one of the nodes in the destination cluster,
what happens if this destination nodes fail?
How can we set XDCR for HA on such case?

Tnx Yaniv

2

Hi yanivg11, XDCR can handle this situation. The replication will resume from the last checkpoint once the destination node comes back online. More details here: https://blog.couchbase.com/understanding-xdcr-part-1/

3

Hi Pavithra,
First thank you for your reply,
Is there a way to set redundancy or some kind of High-Availability for XDCR as well?

Tnx Yaniv

4

There are a few options available. You could set up a bi-directional replication, so source and target clusters will be eventually consistent in case of failures. You could also set up a 3 (or more) cluster topology with 1 or 2 redundant clusters. Finally, you could use backup/restore tools in conjunction with XDCR for disaster recovery.

5

Hi Pavithra,
Thank you once again,
We have 3 different production sites , which we plan for each of those sites to hold a Couchbase cluster built from 3 nodes.
Our goal is make sure that the data is always identical in each of the clusters, and keep them in sync.
Which approach would you think be the most efficient for such scenario?

Tnx Yaniv

6

I would recommend ring topology, Yaniv: cluster 1 -> cluster 2 -> cluster 3 -> cluster 1.

7

Hi Pavithra,
Thank you once again, we would go with your suggestion of the ring topology,
We were also thinking we could set two rings, each in opposite direction tor redundancy of the XDCR:
Ring A: cluster 1 -> cluster 2 -> cluster 3 -> cluster 1.
Ring B: cluster 1 -> cluster 3 -> cluster 2 -> cluster 1.

Is there any prevention for such topology?
Tnx Yaniv

8

Yes, you can go ahead with that…it’s ring topology with bi-directional replications.