support setting password for jupyter notebook

[HelloJeremy]

通过docker-compose部署kubeFATE，怎么设置浏览器访问 jupyter notebook（http://ip:20000）的登录密码
如果jupyter notebook没有访问密码，jupyter notebook会有安全漏洞

[n063h]

通过docker-compose部署kubeFATE，怎么设置浏览器访问 jupyter notebook（http://ip:20000）的登录密码 如果jupyter notebook没有访问密码，jupyter notebook会有安全漏洞

@HelloJeremy We have no such proposal at present, you can manage it manually by

1. enter the client container.
2. generate a hashed passwd {P} by from notebook.auth import passwd;passwd()" with python.
3. ps -aux | grep jupyter to find the args to start the jupyter notebook server.
4. kill the process and rerun the command with replaced --NotebookApp.password={P}

We will appreciate it if you're interested in it and propose a pr to the issue.

[HelloJeremy]

@n063h Jupyter notebook port is 20000，Killing jupyter process triggers client container restart .So this doesn't solve my problem

[HelloJeremy]

@n063h Can NotebookApp.password be designed to be configurable? For example, NotebookApp.password can be specified in parties.conf

[n063h]

@n063h Jupyter notebook port is 20000，Killing jupyter process triggers client container restart .So this doesn't solve my problem

@HelloJeremy Yes. I'm aware of this problem. It's probably related to the entrypoint in dockerfile. I will submit a patch to fix it soon.

[HelloJeremy]

@HelloJeremy Yes. I'm aware of this problem. It's probably related to the entrypoint in dockerfile. I will submit a patch to fix it soon.

@n063h It's fixed, please let me know, thanks

[n063h]

@HelloJeremy Hi, I fixed this issue in PR #790 .If you're using 1.9.0, you may refer to my changes to solve your problem.

[HelloJeremy]

@HelloJeremy Hi, I fixed this issue in PR #790 .If you're using 1.9.0, you may refer to my changes to solve your problem.

@n063h Hi, Does the client docker image need to be updated?

[n063h]

@HelloJeremy Hi, I fixed this issue in PR #790 .If you're using 1.9.0, you may refer to my changes to solve your problem.

@n063h Hi, Does the client docker image need to be updated?

To avoid such different behavior between images of different versions, I choose to write extra entrypoint command rather than modify it. Now the entrypoint in dockerfile has been overwritten in docker-compose.yaml or statefulset.yaml, so you don't need to update the image.

[HelloJeremy]

@n063h Do you need to use $$?

[n063h]

@n063h Do you need to use $$?

I have test it, and it works. I don't think it's necessary. Is there any hidden issue with ${val} ?

[HelloJeremy]

@n063h Hi, I also overridden the entrypoint in docker-compose.yaml, I run /bin/sh -c flow init --ip ${FATE_FLOW_IP}......
But flow init failed，Because I can't get the value of FATE_FLOW_IP, FATE_FLOW_PORT. I'm afraid you have the same problem so suggest you use $$.
You have no problem, Is it because I use /bin/sh you use bash?

[n063h]

@n063h Hi, I also overridden the entrypoint in docker-compose.yaml, I run /bin/sh -c flow init --ip ${FATE_FLOW_IP}...... But flow init failed，Because I can't get the value of FATE_FLOW_IP, FATE_FLOW_PORT. I'm afraid you have the same problem so suggest you use $$. You have no problem, Is it because I use /bin/sh you use bash?

@HelloJeremy Thx for your comment!
I realize it's different behavior between docker-compose and helm-chart. Under testing in k8s mode $ works well, but it doesn't in docker mode.
$ $ is correct one. I will submit a patch to fix that.