Recommendations for upgrading running production API GW

[Marianne Härdh]

We have a two cluster KONG API GW with Postgres as backend and fronted by a load balancer.

We work strictly with automated scripted "infrastructure as code" so patching of running servers is not done. Instead we launch new instances with new code and if they are stable we point the DNS CNAME to the new instances, make them active.

So how can I do an upgrade of the GW keeping the disturbance of API traffic to a minimum? I can launch a new instance with new API version and that will connect to the backend Postgres DB and do the "kong migrations up". Will this affect the running nodes in the cluster, i.e. will they still be able to serve traffic during the upgrade of the schema in the database?

After the database migration is finished, I would then launch two new instances with new API version (without "kong migrations up" command) and then do the switch to make them active.