Reamer commented on a change in pull request #3884:
URL: https://github.com/apache/zeppelin/pull/3884#discussion_r475727402
##########
File path: k8s/interpreter/100-interpreter-spec.yaml
##########
@@ -78,6 +78,21 @@ spec:
- name: spark-home
emptyDir: {}
{% endif %}
+ {% if zeppelin.k8s.interpreter.group.name == "flink" %}
+ volumeMounts:
+ - name: flink-home
+ mountPath: /flink
+ initContainers:
+ - name: flink-home-init
+ image: {{zeppelin.k8s.flink.container.image}}
+ command: ["sh", "-c", "cp -r /opt/flink/* /flink/"]
Review comment:
All in all you are right.
> add different kinds of interpreter related files together may make
interpreter image too large.
Therefore we want to give the users the
[responsibility](https://github.com/apache/zeppelin/pull/3859/files#diff-f55297559468e8c543ae5ab6ee356ec4R66)
to build the best image for their use case.
I think the copy process reduces the start time of the container and it
makes the `interpreter-spec` more complicated. Today we have a special flink
routine, tomorrow another one.
> it’s not flexible to change the interpreter related files，for example, if
we need to change the flink version, we have to rebuild interpreter image
I think this is the biggest advantage and disadvantage of images. If you
need to update any software version, including Flink, OpenJDK and operating
system libraries, you have to rebuild your image. The advantage, you can make a
nice rollback if your image contains all the application software. The
disadvantage, of course, it that you have to rebuild your image, which takes
quite a long time.
If you want to start certain interpreter images in your environment to
reduce the image size, I recommend to create two or more interpreter images and
inset a small if-then-else change
[here](https://github.com/apache/zeppelin/blob/8ab7e9c473c5ce5b487bbab2a7b682f3fb165cab/k8s/interpreter/100-interpreter-spec.yaml#L45).
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
For queries about this service, please contact Infrastructure at:
[email protected]
