2

We have a database cluster built on Percona MySQL. We are going to extend our redundancy with another database cluster with a new data center located in another GEO location. So that in case of failures in one location, we can switch the service to other location.

What is the right technology to keep data in both of GEO sites synced always?

Muneer
  • 146
  • 1
  • 6

1 Answers1

1

You have a 3-node Percona XtraDB Cluster now? And all 3 nodes are in the same DataCenter? And you want to use a second DC for HA?

Perhaps the 'common' way is to set up asynchronous replication (the old Master-Slave) between one of the nodes in the original DCs and a node of a new 3-node PXC cluster in the new DC.

Another approach is to continue to use a single PXC cluster --

  • 4 nodes, 2 in each datacenter.
  • "weight" the nodes so that loss of one DC will let the other DC continue to run. (Caveat: That works one one direction; for full protection from DC failure, you need 3 DCs, with at least one node in each.)
  • Tweak some of the tunables due to the delays between data centers.
  • Set up some sort of Proxy to assist with the automatic routing of queries. Such can probably coexist on the PXC nodes.

(Sorry, there are still more details to ferret out; this is only an overview.)

Rick James
  • 80,479
  • 5
  • 52
  • 119