create DB failed: failed to ping database: dial tcp 127.0.0.1:3306: connect: connection refused

[asm-cygu]

Description create DB failed: failed to ping database: dial tcp 127.0.0.1:3306: connect: connection refused

Reproduction Steps

• run docker (https://github.com/sql-machine-learning/sqlflow/blob/develop/doc/quick_start.md)
• start iris-dnn.ipynb
• execute first cell [ %%sqlflow describe iris.train; }

Expected Behavior cell succed

Screenshots https://raw.githubusercontent.com/asm-cygu/funniestrepo/master/Capture%20du%202020-05-29%2014-45-52.png

Environment (Please complete the following information):

• OS: host centos
• Browser:
• Version:

Additional Notes

[wangkuiyi]

Thank you for reporting the bug! This might be due to a recent change https://github.com/sql-machine-learning/sqlflow/pull/2315. I will look into this bug.

[asm-cygu]

You are welcome ! Thx :-)

[wangkuiyi]

This problem doesn't exist in old versions of the all-in-one Docker images. However, if I docker pull sqlflow/sqlflow and docker run, I got the following error (complete message attached) when running the first SQL statement in the Iris notebook:

[wangkuiyi]

@asm-cygu Sorry for the late reply. This issue is due to that we are working intensively upgrading the SQLFlow deployment.

Currently, the SQLFlow compiler has two modes - the local mode that translates a SQL program into a Python program and the production mode that generates an Argo/Tekton YAML file. It has been a challenge to maintain these two modes and keep them consistent with each other. Considering that all SQLFlow deployments, including those in Ant Financial, DiDi, and Alibaba Cloud, are using the production mode, we want to remove the local mode, at least for the very near future. To make it easy for developers to use SQLFlow, we are building https://github.com/sql-machine-learning/playground, which inherits from the CI setup and release a VM image that includes minikube, Argo on minikube, and all other components. It will run as a single VM, locally by VirtualBox, or remotely on any AWS, Google Cloud, and Alibaba Cloud VM, and it contains all components we use in production deployment.

You are very welcome to watch our progress, especially those PRs tagged playground

[asm-cygu]

I keep watching your great progress closely !

@wangkuiyi seems to said that an old docker container worked, can you upload it to docker hub with a tag ( like sqlflow/sqlflow:lastallinone ) ? You can update your readme too to keep an "easy to start" mode during the development of playground and avoid multiple reports like mine .

Good work again !

[asm-cygu]

or easyer way ... provide me the latest working digestID to pull him :-)

docker pull sqlflow/sqlflow:[Digest]

[wangkuiyi]

What a pity that we didn't yet have tagged (version) Docker images on DockerHub.com. And I brutely overwrote my local one by running docker pull ... Let us work intensively to make the new one and update it here.

Precisely, we do have a versioned one, but I am afraid it is too old (five months) and a little away from our recent work.

[asm-cygu]

Courage and hold on ;-)

[asm-cygu]

unfortunatly v0.2.0-rc.1 is not enought ;-)

root@cb2b7bd3d5ff:/# ps afx PID TTY STAT TIME COMMAND 18 pts/1 Ss 0:00 /bin/bash 28 pts/1 R+ 0:00 _ ps afx 1 pts/0 Ss+ 0:00 bash /start.sh 8 pts/0 Sl+ 0:00 sqlflowserver

[rpjayasekara]

the latest version of the SQLFlow docker image does not include the MySQL because of that we cannot start the MySQL service locally