Thanks for the quick reply.
In our use case, we only use the python and db interpreters (no yarn, spark,
etc). We would start the Zeppelin server in one node (as a container).
Whenever, user runs a notebook, we would spawn a remote interpreter container
(in the same node). We use the isolated mode (one remote interpreter per
note). If we run out of resources in the current node, we would like to spawn
new remote interpreters in a different node (we manage all nodes ourselves).
We are using the DockerInterpreterLauncher to launch the remote interpreter
container.
I have a question about this script (used to start the interpreter inside the
container). I see the script accepts the zeppelin.server.rpc.host, which
allows the interpreter to communicate back to the zeppelin server. If I start
the interpreter in a different node, how does it tell the zeppelin server to
connect to that node? Please advise.
{{CONTAINER_ZEPPELIN_HOME}}/bin/interpreter.sh -d
{{CONTAINER_ZEPPELIN_HOME}}/interpreter/{{zeppelin.interpreter.group.name}} -r
{{zeppelin.interpreter.rpc.portRange}} -c {{zeppelin.server.rpc.host}} -p
{{zeppelin.server.rpc.portRange}} -i {{zeppelin.interpreter.group.id}} -l
{{zeppelin.interpreter.localRepo}} -g {{zeppelin.interpreter.setting.name}}
On Wednesday, June 23, 2021, 11:17:00 AM EDT, Jeff Zhang
wrote:
Yes,it is supported. You can either k8s mode (interpreter runs in pod ) or
yarn mode (interpreter runs in yarn container)
http://zeppelin.apache.org/docs/0.9.0/quickstart/kubernetes.html
http://zeppelin.apache.org/docs/0.9.0/quickstart/yarn.html
denny wong 于2021年6月23日周三 下午10:14写道:
> Hi
> We are using Zeppelin 0.9 and running the remote interpreter as a docker
> container. Currently, the Zeppelin server and the remote interpreter are
> running on the same machine. We would like to know if the remote
> interpreter can be run on a different machine than the Zeppelin server. Is
> this currently supported?
> Thanks
> Denny
--
Best Regards
Jeff Zhang