Cluster, HA and replication

[lumox]

I came across this topic on Reddit:

Compact network and cluster setup

I was wondering if this kind of configuration + HA would make all data perfectly synchronized in case the main node goes down. In particular, let's assume that a mysql container or a VM with it running inside in the node 'one' of the cluster goes down suddenly, would all mysql data (even the last record updated a few seconds before) be immediately available on the other node? Thanks

 

[LnxBil]

In particular, let's assume that a mysql container or a VM with it running inside in the node 'one' of the cluster goes down suddenly, would all mysql data (even the last record updated a few seconds before) be immediately available on the other node?

If this is a real shared storage, then yes.

 

[lumox]

If this is a real shared storage, then yes.

Please define "real" shared storage. Thanks

 

[hec]

NetApp or any other enterprise storage system.

 

[lumox]

NetApp or any other enterprise storage system.

 

[LnxBil]

Please define "real" shared storage. Thanks

Yes, enterprise grade so that you don't have a SPOF. Of course a simple NFS/CIFS/iSCSI NAS would be OK, but why cluster it then (besides experimentation, research and learning)? If the storage failes, everything is down. Just go with one node including storage and do regular backups which has the same failure probability and uses less energy, easier to maintain etc.

The other route would be to drop the dedicated shared storage (NAS) and go with CEPH and build a distributed shared storage system without any SPOFs. That would be the prefered way if you must have perfectly synchronized data (at least crash consistent).