Not the answer you need?
Register and ask your own question!

can we achieve HA with 2 nodes/pods ?

skp1234skp1234 Current User Role Novice
  1. Actually what happening is that with a 2-node cluster, if 1 node goes down the cluster become nonfunctional, is there any way can we transfer incoming requests to another available pod ? or any master-slave concept available with two-node to achieve HA?
  2. is it possible to sync data from one cluster to another in real-time and take the traffic on both clusters?  
  3.  I am a little confused about the PVC of proxySql, will it also consist of the data of the corresponding pod which is pointing to and if we enable proxysql with 2 nodes then will end up with data replications on all 4 PVC?,                                                                                                                                                               I am a newbie to this operator so please bear with me. 
Tagged:

Answers

  • MykolaMykola Percona Percona Staff Role
    1. in PXC you can write to any member anytime, it uses synchronous replication inside. we strictly don't recommend 2 members cluster because of the split-brain issue. The operator has proxy component to handle automatic failover.
    2. async replication between PXC clusters is not supported in the Operator right now.
    3. ProxySQL can be configured via admin port, you can set many different settings manually (like query rules, proxy users, etc), without setting them via Operator. PVC is needed for storing these settings between restarts of ProxySQL.
  • skp1234skp1234 Current User Role Novice
    edited August 25
    many thanks Mykola, I understand this split-brain issue with an even number of nodes, but even with three nodes when it loses its quorum svc becomes unavailable and not able to connect with the database, so no read and write is happening at that time.
Sign In or Register to comment.

MySQL, InnoDB, MariaDB and MongoDB are trademarks of their respective owners.
Copyright ©2005 - 2020 Percona LLC. All rights reserved.