How to easily and elegantly roll-up nats-server instances to the latest version within a cluster

[Doslin]

How to easily and elegantly roll-up nats-server instances to the latest version within a cluster

[Doslin]

How to easily and elegantly roll-up nats-server instances to the latest version within a cluster

The restart command nats-server --signal reload=1 doesn't seem to work

[derekcollison]

Shutdown with lameduck signal.
Start new version server, wait for /healthz to return ok.
Move to next server.

[bruth]

@Doslin For the lameduck signal, you can do nats-server --signal ldm=1 (assuming PID is 1) or send a SIGUSR2 signal to the process.

[bruth]

For the health check, once the new server starts, you can do something like: curl --fail --silent --retry 5 --retry-delay 1 http://my-nats-server:8222/healthz > /dev/null to assert it is healthy. This translates to retry this endpoint 5 times until it returns a 200, otherwise considered fail with a non-zero exit code.

[derekcollison]

This is built in to the helm operator.