tree: e4f89eb4366b321af759fc320c659c09917ff795 [path history] [tgz]
  1. angle/
  2. chrome/
  3. chrome-public/
  4. cros-tast/
  5. eskia/
  6. flutter/
  7. flutter-engine/
  8. koru/
  9. local/
  10. lottie/
  11. lottie-spec/
  12. pdfium/
  13. skia/
  14. skia-infra/
  15. skia-public/
  16. .gitignore
  17. BUILD.bazel
  18. config_files.bzl
  19. goldpushk-corp-test1-instance.json5
  20. goldpushk-corp-test2-instance.json5
  21. goldpushk-test1-instance.json5
  22. goldpushk-test2-instance.json5
  23. README.md
golden/k8s-instances/README.md

Gold instance configurations

json5 files

These are the configurations for each Gold instance. They are used by various Gold services, built into the container, and passed as arguments in the yaml files in the k8s-config repo.

CockroachDB data retention policy files

These are SQL files that need to be executed manually to apply data retention policies on certain tables. We use Row-Level TTL to automatically delete expired data.

Data retention policies are implemented as configurations in the data layer rather than part of the application schema because:

  • The policies may vary for each instance.
  • Gold does not depend on the policies to function, and should not use those columns (expire_at).
  • This separation of concern allows us to evolve the implementation of data retention policies in the future.

Guidelines when defining policies for a specific instance.

  • Keep data at least 2x the size of the commits sliding window.
  • Define proper policies based on the purpose of a table. For example, the ValuesAtHead table is a “caching” table that requires a policy similar to LRU cache; the Expectations table is a “core entity” that should retain all triaged records indefinitely and only delete overdue un-triaged ones.
  • Refer to existing policies of other instances for techniques and consistency.