This message was deleted.
# troubleshooting
s
This message was deleted.
i
looks like it is trying to connect to localhost as the coordinator and not able to. Is the coordinator located on the same node? Does localhost resolve correctly?
a
coordinator is on different server and the query server can reach it (verified by netcat)
i
Was this run previously on zookeeper as a single node cluster?
a
all the servers are single node including zookeeper. The behavior is similar for fresh setup or retires. Have also tried shutting down all servies and deleting the path in zookeeper and bringing up all services.
i
a
yeah I did follow this thread. Couldnt seem to be working for me.
i
check that the router can talk to zookeeper as well
also that the router config has the correct
Copy code
druid.router.coordinatorServiceName
a
there is connectivity between router and zookeeper. Also the router logs convey it.
service name looks correct to me. I hope I am checking at correct location
i
are you running the coordinator under the overlord process?
I should say overlord under coordinator
a
I am not sure, I have not done any other changes explicitly, just kept it the default.
i
do you have a separate process for overlord and for coordinator?
a
no, it is same
logs file is a combined one too
i
the issue is in the discovery process. The router cannot seem to find the coordinator. I would check the paths via the zkCli to make sure there are existing paths as it is configured. It may be because it is being run as one service
a
okay. when i query zk I can see the path for both coordinator and overload
I will try to run them separately and check the behaviour
i
check coordinator log as well to see if any errors stand out regarding zookeeper
a
I do see exceptions in master intermittently when connecting to data server for the ports 8083 and 8091
i
your common.runtime.properties has druid.host=localhost. Try removing that
a
after removing the services started fine except for router. but now all the services are showing this exception
i
I believe what is happening is each service is being registered as localhost. That is now in zookeeper. The default for druid.host is InetAddress.getLocalHost().getCanonicalHostName()
please try that and try cleaning out the zk directories as you did before
a
okay, what you are suggesting is stopping all services, cleaning zk and starting them back (no changes in common.runtime.properties)
i
you need to put in druid.host with the default value, localhost is for single node clusters
druid.host=InetAddress.getLocalHost().getCanonicalHostName()
a
okay let me do that
i
you should just be able to comment out druid.host on all common.runtime.properties and it will use the default value
did you previously just remove the value?
and that is why it was throwing null
a
previous i removed the entire line
i
hmm ok
a
😕
i
I am unsure why it is not working. It should have used the default when you removed from all common.runtime.properties files and cleaned out zookeeper
that is definitely the issue though
you could put the ip or hostname in on each one but I do not believe that is necessary
a
you mean druid.host to their respective ips?
i
yes
a
setting them to their respective ips all services are up and running
no errors so far
🎉 1
druid cluster is up and running
@Ian Roberts Thank you for your time and expertise. I could not have made it work otherwise 👏
i
you are very welcome, glad it is working
🙌 1