Another question. I am deploying pinot via communi...
# troubleshootingp
Priyank Bagrecha
07/26/2022, 11:03 PMAnother question. I am deploying pinot via community provided helm charts, and it seems like the server just fails to restart but I am not sure why.
x
Xiang Fu
07/26/2022, 11:04 PM
I think pinot controller is not yet up, for the first time installation if server starts before controller, then it will fail until controller is fully up and set up all the necessary zk paths.
p
Priyank Bagrecha
07/27/2022, 12:22 AMI see. I thought the helm chart would make sure the components start in the right order.
Priyank Bagrecha
07/27/2022, 12:23 AMHow were you able to deduce that controller is not fully up yet? Is it based on experience or something in the logs was dead give away?
x
Xiang Fu
07/27/2022, 1:01 AM
typically zk should go up first, then controller goes up to set up the helix structure in zk, e.g. create the path of
/<helix-cluster-name>/…Xiang Fu
07/27/2022, 1:02 AM
broker/server start with the assumption that the helix structure is already there, so they may get npe if controller starts slower
Xiang Fu
07/27/2022, 1:02 AM
since this is for cluster setup and first time, so we are not mind to fix the strict ordering
p
Priyank Bagrecha
07/27/2022, 2:13 AMI deleted the server pod so that k8s can start another one in its place, but the new one also had the same issue. Please note that the zookeeper, controller and the broker are already running. Do I need to clear zookeeper entries for the server prod for it to be able to come up?
4 Views