Why does pgo decide on node placement? about postgres-operator

[jmccormick2001]

actually you caught an error on my part which I forgot to change, in cluster-deployment-1.json, I should have specified preferredDuringSchedulingIgnoredDuringExecution instead of requiredDuringSchedulingIgnoredDuringExecution

I'll make that change and push it to master. This will allow the scheduler to move the PG master in the event of a node failure as you would expect.

[jmccormick2001]

but to better answer your question, some people will want to place the master on a specific node over others in certain situations, if you don't specify --node-name on cluster creation, pgo should just pick a node at random for the master. It will attempt to place the replicas on nodes that are not where the master is located.

[tarasglek]

There is another bug. Atm pgo picks first node by default. In my cluster first node is tainted(master node). Feels like a better default behavior would be to not specify a node and let kubernetes scheduler decide.

[jmccormick2001]

thanks, I'll fix this, I agree, the default behavour should not be to specify a node.

[jmccormick2001]

merged to master.

[jmccormick2001]

I plan on another release next week, around Thursday.

[jmccormick2001]

I decided not to wait, and went ahead and pushed version 1.5.2 which should fix this problem, thanks for reporting it.

[jmccormick2001]

should be fixed now.