Discover Cells 1.0 impact for project_management

Description

Hello team, as many of you already knew, the Cells project is one of the top priorities for FY2025, with the goal of providing additional scalability for GitLab.com. It is a very large and highly complex project that will take multiple iterations. As the first iteration, the Cell project team proposed the design of Cells 1.0, with a focus on new enterprise customers. This iteration will allow us to get hands-on experience deploying and operating Cells sooner and be a big step towards our ultimate goal of alleviating scalability pressure on gitlab.com SaaS.

The Cells 1.0 design and execution plan were discussed in the Engineering FY23Q4 Offsite and received approval. Our plan is to have it in production by mid of FY 2025.

Cell 1.0 requires the whole company to participate in this project and make respective features and workflows compatible with Cells 1.0.

Below are the details that are helpful for all engineering groups to start looking at their technical domains.

What we are asking for

We are asking all product teams to take the following steps:

Follow the guidelines in this Discovery Epic to see how the Cells 1.0 architecture affects your domain:
- Review Cells 1.0 design, and identify the workflows in your product area that would be affected by Cells 1.0.
- Test your workflows against the Cells 1.0 POC.
- There is an issue for each team in the Discovery Epic. You can use this issue to discuss the implications of Cells 1.0 with your group, or involve Tenant Scale engineers if you need more clarification.
- If you discover workflows that need to be fixed to be compatible with Cells 1.0, open sub-epic for the workflows in the Cell 1.0 Epic and break down the effort into issues.
- Estimate efforts (T-shirt sizing, we are NOT asking for a detailed design and work estimate), and provide a high level timeline (take into account resource allocation) in the sub-epic created above. We know there are a lot of unknowns, so try your best to give a rough initial estimation. The goal here is to allow us to identify risk areas and address them early.
Add sharding keys to all your cell local tables, see the documentation and progress dashboard.
- All tables with the following gitlab_schema are considered “cell-local”:
  - gitlab_main_cell
  - gitlab_ci
Classify your tables under the different database schemas (gitlab_main_cell or gitlab_main_clusterwide), see the documentation and progress dashboard.

We are working to automate adding the sharding keys (2) and classifying the tables (3), but we won’t be able to automatically classify everything. So if you want to be proactive, we won't need to contact you again in case we can't automate the work.

The Cell Project team will work out an overall project timeline with a target end of July production date. Please take into account resource allocation when doing your timeline estimation.

We are also preparing an AMA to explain the details of the ask.

If your team already completed a Cells discovery issue before this, we still request that you re-evaluate your workflows with the Cells 1.0 scope in mind.

Tips for Discovery

Read the Cells 1.0 design for the plan of the first iteration. Ask the following questions and discover incompatibilities in your features following the clue from the answer -

Does your team use or require admin accounts on GitLab.com?
Does your team require direct database access in a form of Rails console, database replica?
Does your team have a service that it manages that is not part of Omnibus or Cloud Native deployment?
Does your team provide instance-wide features in the product? Ex.: shared runners, GitLab Pages, etc.
Does your team use or implement authentication features? Ex.: login via access token, Deploy Tokens, Kubernetes Agent connecting to GitLab.

Relevant links

Cells 1.0 Executive Summary
Cells 1.0 Architecture Design
Discovery Epic - follow instruction in it to discover if any work needed to make your product compatible with Cells 1.0
Cells 1.0 Epic - create sub-epics and issues to track work you identified
Cells 1.0 workstreams slides and video walkthrough

Timeline of Immediate Tasks

2023-01-31 - Two AMAs (APAC/EMEA and AMER/EMEA friendly) and PoC is sent out to teams.
2023-02-06 - Due date for all engineering groups (development, infrastructure, test platform) finish breakdown and work estimate.
2023-02-09 - First version of Cell 1.0 Schedule finished and communicated, with the understanding that we may have more iterations in the future.