Add scripts for create ubuntu jammy docker file

[leborchuk]

In these PR I added to workflow building jammy Docker container.

The main peculiarity of the Debian package is that we need to build the binary in a specific directory and then install it in /usr/local/cloudberry-db. This differs from the package creation process for RPM-based distros, where you simply list what you want to package (*.spec) along with absolute paths. For Ubuntu you should list relative local build paths and then write install script to copy it in /usr/local/cloudberry-db.

So, I added the ${BUILD_DESTINATION} path (with defaults to /usr/ local/cloudberry-db) to all build automation scripts. This has made my PR slightly more dangerous, because if something goes wrong (I don't think I have, but it needs to be thoroughly reviewed), all our builds could suffer.

[leborchuk]

@edespino Could you please review my PR? I want to add deb packages to cloudberry and this is my first PR where I try to create Docker images.

[tuhaihe]

Cool @leborchuk, thanks for your great work. I also would like to help review this PR.

[leborchuk]

I found out that the deb package's dependencies included our internal python packages, so removed them.

[tuhaihe]

Left some non-technical comments, FYI. I will try to test this workflow in my environment, and if any issues, I will let you know. Thanks!

[leborchuk]

I fixed issues, PR can be reviewed

[tuhaihe]

Hey @leborchuk , an error returned when running docker build -t cloudberry-db-base-env . under the dir images/docker/cbdb/test/ubuntu22.04:

Seems some packages cannot be installed from the ubuntu 22.04 software sources:

glibc-locale-source ~> glibc-source
openssh
openssh-clients ~> openssh-client
procps-ng ~> procps
which

[tuhaihe]

After building and installing Cloudberry in the build image, I run the test for twice with the following command:

PGOPTIONS='-c optimizer=on' make installcheck

Some errors return:

regression-diffs.txt

 +20250401:07:07:21:037248 gpconfig:cdw:gpadmin-[ERROR]:-USER environment variable must be set.
+USER environment variable must be set.

Should we add the ENV USER=gpadmin to the build Dockerfile? FYI. Will test later.

[edespino]

APT Flags & Repos

Problem: You're using --allow-unauthenticated and Acquire::AllowInsecureRepositories=true. This is not safe for production.
Fix: Unless absolutely required (e.g., for local mirrors), remove these flags. If used, explain the context in a comment.

[edespino]

Duplicate Packages

libssl-dev appears twice in the APT list.
libaprutil1-dev and libuv1-dev are also later reinstalled under gpadmin.

Fix: Install once globally under root and avoid duplication unless version/context differs.

[edespino]

Missing Cleanup

Problem: The APT cache and downloaded files (e.g., .tar.gz, build dirs) are not cleaned up, which bloats the image.
Fix:

RUN apt-get clean && rm -rf /var/lib/apt/lists/*

Also clean /tmp/, ~/xerces-c, Go tarballs, etc., after use.

[edespino]

Timezone hardcoding

ln -snf /usr/share/zoneinfo/Europe/London is hardcoded, even though you have TIMEZONE_VAR.

Fix: Use the variable:

RUN ln -snf /usr/share/zoneinfo/${TIMEZONE_VAR} /etc/localtime && echo "${TIMEZONE_VAR}" > /etc/timezone

[edespino]

Repetitive locale settings

You run locale-gen and update-locale multiple times.
Fix: Consolidate to a single layer.

[edespino]

Have you considered slimming the image using ubuntu:jammy-minimal?

[edespino]

I have submitted several comments already. I have more to review.

[yjhjstz]

Great works~~

[tuhaihe]

Hi, there is another issue to resolve: the /tmp/init_system.sh cannot be executed when the new image starts up. FYI.

[edespino]

@leborchuk This is great progress — thanks for putting in the effort to bring Ubuntu 22.04 support to Cloudberry!

That said, I’d really prefer to see this rolled out in more focused, logical chunks so each stage is easy to test, review, and validate independently. Something like:

1. Developer workflow – Introduce base support for Ubuntu 22.04: container build image, compiler config, developer installcheck support. A developer should be able to build and run installcheck against main.
2. CI build automation – Add a GitHub Actions workflow to build the Ubuntu 22.04 developer container image automatically.
3. Release engineering workflow – Extend the container image to support .deb package creation. A developer should be able to build and produce a .deb file inside the image.
4. QA workflow – Introduce a minimal runtime container image (based on Ubuntu 22.04) to install and validate the .deb package. This image should support basic Cloudberry functionality and simulate a QA-style install, including gpadmin setup and dependency resolution.

Splitting the PR this way would make it easier for others to review and test each part without merging everything at once. Would you be open to restructuring the work along these lines?

[leborchuk]

Hi, @tuhaihe@edespino I want to reopen PR once again.

I see that no visible commits have been made to https://github.com/edespino/cloudberry/tree/ubuntu-22.04, so I decided to move forward.

I took all the work that had been done by @edespino and set up my own environment for creating Debian packages. Then, I created a Docker environment and Debian package using it (using github actions). I tried to install the Debian package in my development environment - everything worked well. Of course, all packages include PAX storage. I also checked if the simple PAX scripts worked on Ubuntu - there were no errors, everything was great.

It's a first PR where I want to add scripts do cloudberry-devops-release repo.

The plan for future work:

1. run workflow and upload created ubuntu docker files to registry
2. create test debian package workflow (manual run) build based on https://github.com/leborchuk/cloudberrydb/blob/AddJammyBuild/.github/workflows/build-deb-cloudberry.yml and check if all works well
3. Add build & test debian packages to https://github.com/apache/cloudberry/blob/main/.github/workflows/build-cloudberry.yml
4. Discuss whether we want to be able to run tests using a special tag on a Debian image or not.

Here you can find details about the workflow and produced artifacts:

1. Dockerfile - https://github.com/leborchuk/cloudberry-devops-release/actions/runs/15562547058/job/43818330748
2. Debian package - https://github.com/leborchuk/cloudberrydb/actions/runs/15662457461/job/44122150152

[tuhaihe]

Thanks @leborchuk for helping push this big feature forward again! Just left some comments, FYI. Will help test earlier this week.

[leborchuk]

Thanks @leborchuk for helping push this big feature forward again! Just left some comments, FYI. Will help test earlier this week.

Thanks, fixed

[leborchuk]

Tip: add here libssh2-devel after merging #27 (review)

[leborchuk]

Rebased to the new cloudberry-env file

[tuhaihe]

Hey @leborchuk, would you like to start planning the migration of this repo to the main one, as discussed here: https://lists.apache.org/thread/p6otyrrnosg8fsbyr6ok7hl8wxpx4ss2? I think it would be better to open a new PR on the main after this repo's migration is complete. Otherwise, it could make the migration more difficult.

How about this?

[leborchuk]

Hey @leborchuk, would you like to start planning the migration of this repo to the main one, as discussed here: https://lists.apache.org/thread/p6otyrrnosg8fsbyr6ok7hl8wxpx4ss2? I think it would be better to open a new PR on the main after this repo's migration is complete. Otherwise, it could make the migration more difficult.

How about this?

Well, do you think the time has come? Let's start it then) I will store scripts from this PR in my own repo, move this repo and fix building scripts, then open PR again

[tuhaihe]

Hey @leborchuk, would you like to start planning the migration of this repo to the main one, as discussed here: https://lists.apache.org/thread/p6otyrrnosg8fsbyr6ok7hl8wxpx4ss2? I think it would be better to open a new PR on the main after this repo's migration is complete. Otherwise, it could make the migration more difficult.
How about this?

Well, do you think the time has come? Let's start it then) I will store scripts from this PR in my own repo, move this repo and fix building scripts, then open PR again

I think it's good to start the plan and move forward now. We can have a discussion about it and see how to reorganize these files in the main repo, then we can take actions. Sounds good?

[leborchuk]

Created apache/cloudberry#1292

[leborchuk]

I created apache/cloudberry#1359 in the main repo, these repo will be archived soon.