Geo Replicate on 2 datacenters

[xsys...@gmail.com]

Hello

Currently I have 4 Servers in my galera cluster at 2 datacenters ,
each datacenter has two servers but when we face network issues on any Datacenter  cluster splits in 2 parts each 2 servers so we have 2 non primary components .
I want to setup 2 new nodes in each datacenter means we have 6 nodes in cluster and if we go with problem on any datacenter we have 3 servers on each dc I think its solve the problem
the question is 6 is even number is there any issues i must expect for this setup ?

[Joe O]

You will still face a split brain decision. You probably need 2 more nodes in a 3rd DC to avoid this. 

--
You received this message because you are subscribed to the Google Groups "codership" group.
To unsubscribe from this group and stop receiving emails from it, send an email to codership-tea...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Arash Shams]

unfortunately I have only 2 locations , what is your suggestion about 2 locations clustering how many nodes per datacenter ?
 

[Alexis Guajardo]

Whats your purpose of having multiple data centers?

With only two data centers,youd have to use weighted quorum, or add a third DC.

Google cloud platform ( GCP ) and Amazon Web services ( aws ) both make it super easy to add a 3rd DC .

On Nov 26, 2016 3:07 AM, "Arash Shams" <xsys...@gmail.com> wrote:

unfortunately I have only 2 locations , what is your suggestion about 2 locations clustering how many nodes per datacenter ?
 

--
You received this message because you are subscribed to the Google Groups "codership" group.

To unsubscribe from this group and stop receiving emails from it, send an email to codership-team+unsubscribe@googlegroups.com.

[Lammert Bies]

With two data centers and an equal number of servers you will run into split brain conditions as soon as the data centers fail to communicate with each other. Adding weighted quorum doesn't help, because the chance is 50% that the cluster in the data center which has the connection failure is actually selected as the cluster with the highest quorum. And then you have no service for your clients anymore, even though part of the servers are still available.

If you don't have the option to add an extra database node in a third data center, you can use a Galera Arbitrator. This is a server which participates in the quorum selection, but doesn't store any data. It does receive all data however as it is part of the replication process, so bandwidth and security should be sufficient.

If this arbitrator server is located at a third location independent of the other two data centers, which could even be your office location, it can help determine which data center is currently off line and mark the remaining cluster as primary.

More information about an arbitrator can be found here: http://galeracluster.com/documentation-webpages/arbitrator.html

You may also want to look at this how to at the Percona website: https://www.percona.com/doc/percona-xtradb-cluster/LATEST/howtos/garbd_howto.html as it describes how to run an arbitrator on a server.