I have three Graylog nodes in a multi-node setup.
One is configured as master. Two not as master.
The Graylog servers are in a manual patch/upgrade window for os updates (every one on another day)
Sometimes a restart is required, which will be done automatically.
I experienced now: If the Graylog leader node is updating and restarting the pipeline is not working at all and the lgos are not processed an end up in the default stream.
So my conclusion is, pipelines only work, if the leader is online (available).
My question is, how would I ensure, that the pipeline gets processed anway? Do I need to change the leader node manually before the patch day?
What does that mean for updating Graylog version itself? If I upgrade the Graylog leader node, the messages will also not be processed in the pipeline. Do I change another node to leader, before I want to update the initial leader, to prevent loosing the pipeline step?
Does “pause message processing” help in this case?
Thank you for the answers.