I have 2 kubernetes node with cockroach pod on each. I’m running a script which makes select from table. When I delete a pod ‘cockroach-cockroachdb-1’ the script always returns a result; when I delete a pod ‘cockroach-cockroachdb-0’ the script doesn’t return a result some time (while kubernetes restart pod).
I don’t understand whether there is main cockroach node which must work always? I assume that all nodes are equivalent to each other and if any node is down it doesn’t influence on productivity.
CockroachDB requires at least 3 nodes in a cluster in order to withstand node failures. This is because it uses a quorum-based replication system. If you add a third node, your cluster will start being able to handle failures.
For a demonstration of this behavior, see https://www.cockroachlabs.com/docs/stable/demo-fault-tolerance-and-recovery.html