MySQL/MariaDB Cluster management software

InceptionHosting

Aside from cli is anyone aware of any sql cluster management software?

Back in my windows admin days, we used to use Symantec cluster manager (I think) for adding failover servers, masters, front ends etc and it had a great management panel for looking at what was causing issues and general performance metrics.

Granted it likely cost a million dollars and was bespoke to some degree but I am struggling to find anything similar (paid or free) for MySql, anyone got any suggestions or knowledge to throw at me?

Cheers.

Ant.

Andrei

Have you tried this? https://severalnines.com/product/clustercontrol/clustercontrol-community-edition

I've been meaning to play with it for quite some time now but haven't gotten to it yet... but it looks very good.

tetech

Another option, https://www.percona.com/software/database-tools/percona-monitoring-and-management

InceptionHosting

@Andrei said:
Have you tried this? https://severalnines.com/product/clustercontrol/clustercontrol-community-edition

I've been meaning to play with it for quite some time now but haven't gotten to it yet... but it looks very good.

I looked at it but the community edition is not self-healing which kind of sucks.

Neoon

@AnthonySmith said:

@Andrei said:
Have you tried this? https://severalnines.com/product/clustercontrol/clustercontrol-community-edition

I've been meaning to play with it for quite some time now but haven't gotten to it yet... but it looks very good.

I looked at it but the community edition is not self-healing which kind of sucks.

MariaDB is selfhealing, until a certain point.
Do you expect one of the DC's to be nuked away by a bomb?

If you keep a decent amount of nodes like 5 instead of the minimum 3, you need 2 failures before things get pissy.

InceptionHosting

@Neoon said: MariaDB is selfhealing, until a certain point.

I am talking about a self-healing cluster, mariadb cant heal itself if the server that it is running on gets nuked

Andrei

@AnthonySmith said: I am talking about a self-healing cluster, mariadb cant heal itself if the server that it is running on gets nuked

Galera cluster can self-heal, as long as 50%+1 of the nodes are still operational.

They recommend running a minimum of 5 nodes per cluster, so you can have 2 servers nuked and still have a healthy and self-healing cluster.

However, I wouldn't recommend Galera if you are running a very actively written database, as the cluster is not fully asynced so nodes would have to wait one after another to confirm every write.

Neoon

@Andrei said:

@AnthonySmith said: I am talking about a self-healing cluster, mariadb cant heal itself if the server that it is running on gets nuked

Galera cluster can self-heal, as long as 50%+1 of the nodes are still operational.

They recommend running a minimum of 5 nodes per cluster, so you can have 2 servers nuked and still have a healthy and self-healing cluster.

However, I wouldn't recommend Galera if you are running a very actively written database, as the cluster is not fully asynced so nodes would have to wait one after another to confirm every write.

As Andrei said, as long you have 3 servers and 2 go off, you can bring them back, they rsync and all good.
But if you end up loosing 3 of your 5, its borked.

imok

Our low end databases won't be self-healing, but at least we are.

tetech

@Andrei said:

@AnthonySmith said: I am talking about a self-healing cluster, mariadb cant heal itself if the server that it is running on gets nuked

Galera cluster can self-heal, as long as 50%+1 of the nodes are still operational.

They recommend running a minimum of 5 nodes per cluster, so you can have 2 servers nuked and still have a healthy and self-healing cluster.

However, I wouldn't recommend Galera if you are running a very actively written database, as the cluster is not fully asynced so nodes would have to wait one after another to confirm every write.

We use it on 5 nodes with about 10 writes/sec and it is OK for our particular profile. For us the I/O time dominates any cluster latency. We additionally run an async slave.

hey

@Andrei said:
They recommend running a minimum of 5 nodes per cluster, so you can have 2 servers nuked and still have a healthy and self-healing cluster.

We use 3 server on production , so far never had problem with data sync when 2 server down.