Automating schema migration flow with GitHub Actions, skeema & gh-ost

Schema migration is more than running an ALTER TABLE. It is about designing, reviewing, approving, queuing, scheduling, executing, auditing, controlling and versioning the changes.

At GitHub we run multiple migrations per day, and much of this flow used to be manual, taking a significant toll from the databases team. In this session we illustrate how we automated away migration using free and open source solutions, and based on trusted development flow.

We highlight the use of the skeema tool, with GitHub Actions, git flow and gh-ost.

Transcript

1. Automating Schema Migrations with GitHub Actions, skeema & gh-ost Shlomi Noach GitHub FOSDEM 2020
2. About me @github/database-infrastructure Author of orchestrator, gh-ost, freno, ccql and others. Blog at http://openark.org 
 github.com/shlomi-noach
 @ShlomiNoach
3. GitHub
 Built for developers Busy and growing 40M+ developers
 3M organizations
 44M repositories created in the past year Actions, Packages, 
 Security Advisories & Updates, 
 Code Navigation & Search,
 Notifications, Sponsors,
 Mobile, …
4. Incentive With MySQL as the backend, new features imply schema changes: new tables, new columns, index changes, iterative schema changes, experiments.
5. What’s in a migration? More than CREATE, ALTER or DROP TABLE
6. What’s in a migration? Designing, coding, local testing, review process, queueing, scheduling, executing, controlling, auditing, versioning…
7. Ownership Developer
 Developer
 Peer
 DBA
 DBA
 DBA
 DBA
 DBA
 DBA/SRE
 DBA
 DBA
 DBA
 Developer
 Developer Design code + schema change
 Publish
 Review
 Review
 Formalize statement/command
 Locate
 Schedule
 Run
 Audit/monitor/control
 Cut-over/complete migration
 Cleanup
 Notify
 Deploy
 Merge
8. Ownership, our previous state Developer
 Developer
 Peer
 DBA
 DBA
 DBA
 DBA
 DBA
 gh-ost/chatops
 DBA
 DBA
 DBA
 DBA
 DBA Design code + schema change
 Publish
 Review
 Review
 Formalize statement/command
 Locate
 Schedule
 Run
 Audit/monitor/control
 Cut-over/complete migration
 Cleanup
 Notify
 Deploy
 Merge
9. A complex flow Multiple domains (code, MySQL, production, communication). Multiple environments (dev, production). Multiple owners (devs, DBAs, SREs).
10. A combinatory solution Loosely coupled, independent components. Each solves an aspect of the problem. Orchestrated to create an automated flow.
11. Code A schema change should be presented as code. Coupled with application code. Versioned.
12. Code At GitHub, we use git. We also happen to author GitHub for code hosting, versioning and management.
13. Pull Request The change Review CI Discussion
14. Code:
 New PR
15. Code: 
 New PR
16. What’s the migration? From code to SQL statement
17. skeema https://www.skeema.io/ https://github.com/skeema/skeema Open source Developed by Evan Elias
18. schema/
 .skeema
 my_schema1/
 .skeema
 some_table.sql
 another_table.sql
 my_schema2/
 .skeema
 foo.sql
 bar.sql
19. $ cat .skeema
 
 [skeema-diff-ci]
 host=127.0.0.1
 port=3306
 user=root
20. $ cat .my_schema1/some_table.sql
 
 CREATE TABLE `some_table` (
 `id` int(10) unsigned NOT NULL AUTO_INCREMENT,
 `hostname` varchar(128) NOT NULL,
 `time_created` datetime NOT NULL DEFAULT CURRENT_TIMESTAMP,
 `time_updated` datetime DEFAULT NULL,
 `random_hash` char(40) CHARACTER SET ascii DEFAULT NULL,
 PRIMARY KEY (`id`)
 ) ENGINE=InnoDB DEFAULT CHARSET=utf8;
21. $ skeema push env-name
 # Connects to MySQL server, applies filesystem schema onto database $ skeema pull env-name
 # Imports schema definition from MySQL onto filesystem 
 $ skeema diff env-name
 # Prints schema migrations changing the state of MySQL server to filesystem

22. skeema Where?
23. GitHub Actions Kick off workflows with GitHub events like push, issue creation, or a new release. An action runs in a container on GitHub’s infrastructure (default). Action has repository’s context and can operate on the repository.
24. GitHub Actions Run skeema from within Action. Fetch skeema as part of Action flow.
25. GitHub Action: skeema-diff, simplified skeema-diff:
 runs-on: ubuntu-latest
 steps:
 - uses: actions/checkout@v2
 with:
 ref: master
 - name: push master schema to MySQL
 env:
 MYSQL_PWD: root
 run: |
 skeema push skeema-diff-ci
 - uses: actions/checkout@v2
 - name: skeema diff
 skeema diff skeema-diff-ci --allow-unsafe
26. GitHub Action: skeema-diff - uses: actions/checkout@v2
 with:
 ref: ${{ github.event.pull_request.base.sha }}
 - name: push master schema to MySQL
 …
 - uses: actions/checkout@v2
 with:
 ref: ${{ github.event.pull_request.head.sha }}
 - name: skeema diff
 /tmp/skeema-ci/skeema push skeema-diff-ci --allow-unsafe --ddl- wrapper='echo "\n-- skeema:ddl:begin\n"{DDL}";\n-- skeema:ddl:end"' | sed -e 's/^USE /-- skeema:ddl:use /g' | sed -n '/^-- skeema:ddl:use /p;/^-- skeema:ddl:begin/,/^-- skeema:ddl:end/p' | tee /tmp/skeema-ci/skeema- diff.sql
27. Action/
 skeema
28. Action/
 skeema
29. gh-ost GitHub’s online schema migration tool. Low-impact (nearly no-impact) in production. Auditable, configurable, controllable. Open source https://github.com/github/gh-ost/
 https://github.blog/2016-08-01-gh-ost-github-s-online-migration-tool-for-mysql/
 https://speakerdeck.com/shlominoach/githubs-online-schema-migrations-for-mysql
30. skeefree Name coined by Tom Krouper A service to orchestrate the flow: PR, Actions/skeema, gh-ost. Developed internally at GitHub
31. skeefree Probes and detects schema change PRs Analyzes skeema changes Initiates and follows up on review/approval status Schedules the migration Runs the migration (gh-ost/direct) Follows up and reports on PR
32. Flow: 
 New PR
33. Flow: 
 New PR
34. Flow: 
 CI build
35. Flow: 
 skeema
 analysis
36. Flow: 
 review
 & label
37. Flow: 
 skeefree analysis
38. Flow: 
 Review request
39. Flow: 
 migration
 execution
40. Flow: 
 deploy
 & merge
41. Ownership: skeefree Developer
 Developer
 Peer
 DBA
 skeema/CI
 skeefree
 skeefree
 skeefree
 gh-ost/chatops
 skeefree
 GC
 skeefree
 Developer
 Developer Design code + schema change
 Publish
 Review
 Review
 Formalize statement/command
 Locate
 Schedule
 Run
 Audit/monitor/control
 Cut-over/complete migration
 Cleanup
 Notify
 Deploy
 Merge
42. Impact Database team work reduced to minutes per week. Developers have visibility into status. Get notified on their PR. Better time utilization; migrations start executing as soon as possible, not based on a human availability.
43. skeefree Uses internal services in GitHub’s infrastructure • Inventory service • MySQL discovery service • Chatops integration • Internal libraries (e.g. logging)
44. Open Source skeefree is coupled with GitHub’s infrastructure: - Inventory service
 - MySQL discovery
 - Chat/chatops We nonetheless hope that the community finds it useful and are releasing it in partial state. Release to be announced.
45. Questions? github.com/shlomi-noach @ShlomiNoach Thank you!