Hi,
There are several ways of doing this, but as long as you have 2 gateways with Master Master replication then you can do it as follows (High level instructions).
- Ensure replication is in sync
- Point the 2nd Gateway at its own localhost DB (To ensure no loss of requests temporarily take it out of the load balance pool to allow restart etc)
- Mask the 1st node from the load balancer and upgrade it along with it's local database.
- Bring the 1st node back into the load balance pool and mask the 2nd node.
- Upgrade the 2nd node
- Point the 2nd node back to the 1st Gateway with itself as failover (as originally set), restart
- Bring the 2nd node back to the pool again.
- Fix replication
Split - Upgrade - Join
You now have both nodes upgraded with zero downtime.
Apiida can help ensure this is done with zero downtime if neccessary. Feel free to get in touch.
Regards
Vince Baker
Principal Architect
Apiida AG