[Feature] Support build docker image from Dockerfile (Task Summary)

[imbajin]

Apache HugeGraph DockerHub: https://hub.docker.com/r/hugegraph/hugegraph

Usage:

# 0. download image (latest)
docker pull hugegraph/hugegraph

# 1. start container
docker run -itd hugegraph/hugegraph  # start server in daemon mode

# (Optional configs to expose network port outside)
docker run -itd --name=graph -p 18080:8080 hugegraph/hugegraph

# 2. (Optional) enter the container & start hugegraph-hubble
docker exec -it graph bash
xxx 

Feature Needs

1. Add Dockerfile for running HugeServer
2. Users can pull/run the image from dockerhub directly.

And anyone who has other suggestion for docker image can comment here

After PR, will update the doc 📝

Update in 2023.11.09:

Due to the project will move to Apache Foundation, we'll use 1.0.0 as the server version & bind hubble with it, reopen this issue until it meets the new need, any suggestions for Dockerfile are welcome.

• [ ] Dockerfile start cmd & restart problem #2343
• [ ] use docker-slim to slim the image
• [ ] use docker-compose to manage computer, it needs to contain etcd and hugegraph-server
• [ ] use DockerHub(autobuild), currently it's not free to use (Submit a OSS request)
• [x] fix that hubble can't use localhost/0.0.0.0 & add a script to update the image
• [x] use a script to run server & hubble together OR use docker-compose to manage them
• [x] support Cassandra(Docker) as backend (compose with HugeGraph) #2307
• [x] keep the tags same with server (like hugegraph/hugegraph:1.0.0) https://github.com/hugegraph/actions/pull/1
• [x] pre-load some data or graphs in container so that users can traverse the graph with one step #2241
• [x] use hubble instead of studio to support new features
• [x] support ARM env (for M1 laptop, refer doc)
• [x] consider better way (like add a option/program to run backward) to avoid modify start-script #2118
• [x] adopt v0.12 (support raft based replication)
• [x] use github actions to publish images
• [x] use jdk11 for runtime #2118
• [x] use multi-stage Dockerfile way to slim the image
• [x] support build from the (latest) source code #2118

Refer:

1. Reduce the size of container images with DockerSlim
2. One docker container with multi process (Not graceful:)
3. How to use dumb-init to replace the pid1 in docker

[javeme]

@imbajin Thanks. do we need a repository like this: https://hub.docker.com/repository/docker/hugegraph/hugegraph

[imbajin]

@imbajin Thanks. do we need a repository like this: https://hub.docker.com/repository/docker/hugegraph/hugegraph

Yep, and it should be associated with the current github repository, then add access token for dockerhub to auto get the change from Dockerfile & autobuild the image

Refer to :

1. Link to Github
2. Autobuild from Dockerhub

[javeme]

@imbajin done

[T-baby]

什么时候支持自动发布到 dockerhub

[imbajin]

什么时候支持自动发布到 dockerhub

如果没有紧急需要的话, 应该是等分布式 rocksdb 版稳定之后一起处理把, 不然要改动的地方还挺多

[T-baby]

下版本的 hugegraph 支持分布式是吧。

[imbajin]

下版本的 hugegraph 支持分布式是吧。

@T-baby 没有特殊情况是哒, 不过就算分布式 rocksdb 没有 12 版发布, 也会重新出一个正式的 docker 镜像(包括 hubble 的)

[niconical]

Really meaningful work, I hope I can apply for this project on GLCC 2023!

[niconical]

It seems hard to find relevant PRs for use github actions to publish images.

[imbajin]

It seems hard to find relevant PRs for use github actions to publish images.

due to the permission lack, we have to publish it in separate repo: (otherwise we can't set the SECRET_TOKEN')

• https://github.com/imbajin/nocode/blob/master/.github/workflows/server-docker.yml
• https://github.com/hugegraph/hugegraph-actions/tree/main/.github/workflows

---

some refer to the auto-build in dockerhub with ASF projects, try these ways:

1. apply the GitHub admin role (just include these repos) to ASF(Apache Software Foundation) to see if they could give us the permission (little probability)
2. create a new org & fork all the needed repos under apache, then update manually to trigger the hook in my fork repos (seems achievable, but it's really clumsy/trick)
3. read the ASF doc or submit a ticket on jira to them to find the help

[DoodleBears]

• [ ] fix that hubble can't use localhost/0.0.0.0 & add a script to update the image

What should I do if I user Docker to run hugegraph/hugegraph and hugegraph/hubble and encounter

服务 localhost:8080 不可用

[imbajin]

• [ ] fix that hubble can't use localhost/0.0.0.0 & add a script to update the image

What should I do if I user Docker to run hugegraph/hugegraph and hugegraph/hubble and encounter

服务 localhost:8080 不可用

could use the real ip like 10.xx.xx.7 directly @DoodleBears

@coderzc ，please assign this bug for your stu~

[coderzc]

@aroundabout PTAL.

[aroundabout]

here

[aroundabout]

• [ ] fix that hubble can't use localhost/0.0.0.0 & add a script to update the image

What should I do if I user Docker to run hugegraph/hugegraph and hugegraph/hubble and encounter

服务 localhost:8080 不可用

Can you provide more information? Like the command you use.

[imbajin]

• [ ] fix that hubble can't use localhost/0.0.0.0 & add a script to update the image

What should I do if I user Docker to run hugegraph/hugegraph and hugegraph/hubble and encounter

服务 localhost:8080 不可用

Can you provide more information? Like the command you use.

@aroundabout @DoodleBears better open a new issue in toolchain to handle it, this issue used for task summary (link other issue/PR), keep it clean (I'll hide them later), thanks~ 🆒

[DoodleBears]

opened: https://github.com/apache/incubator-hugegraph-toolchain/issues/490#issue-1796657832

[DoodleBears]

• [ ] fix that hubble can't use localhost/0.0.0.0 & add a script to update the image

What should I do if I user Docker to run hugegraph/hugegraph and hugegraph/hubble and encounter

服务 localhost:8080 不可用

Can you provide more information? Like the command you use.

mentioned here: https://github.com/apache/incubator-hugegraph-toolchain/issues/490#issue-1796657832