Thanks @hswimelar - just some minor questions from me.
I do not have much to add from Organization/Tenant scale perspective
I am confused here. Are we talking about the monolith namespace, or the artifact registry namespace ?
Offtopic: Is numeric ID the right datatype to use for Modular modules ?
Thought experiment. If Modular modules are deployed per cell. Every cell-local Artifact Registry starts the sequence with 1. We will get ID clashes.
Hi @tskorupa-gl can you please post a brief update here ?
All
oauth_applicationshave a value fororganization_idin the .com DB.
Nice ! Thanks for the detailed update @jknabl-gitlab
Prompt: Generate specs for ee/app/models/analytics/aggregation_engines/finished_pipelines.rb. Local operations only.
Branch: !226711, without this commit
Result: Generated ee/spec/models/analytics/aggregation_engines/finished_pipelines_spec.rb in repeated lets.
Branch: !226711 rebased on top of master containing this commit.
Result: Generated ee/spec/models/analytics/aggregation_engines/finished_pipelines_spec.rb in repeated lets.
Branch: !226711 rebased on top of master containing this commit. And .ai/testing.md to read doc/development/testing_guide/best_practices.md.
Result: Generated ee/spec/models/analytics/aggregation_engines/finished_pipelines_spec.rb in repeated lets.
The generator seems to be heavily weight patterns found in existing specs.
→ Read 3 files ↳ ee/spec/models/analytics/aggregation_engines/agent_platform_sessions_spec.rb ↳ ee/spec/models/analytics/aggregation_engines/code_suggestions_spec.rb ↳ ee/spec/models/analytics/aggregation_engines/contributions_spec.rb
Looks like we have some way to go to encourage agents to use table-based tests.
LGTM @tkhandelwal3 - just a minor update suggestion from me.
note: I was not 100% about using minval > sequence.last_value as guide. Here's one possible scenario that Duo generated:
A new cell sequence that happens to have been used a little: If a new cell's sequence has been called even once (e.g., last_value = 5), and minval = 1, then minval > last_value is false, and it would be treated as a legacy cell — incorrectly skipping the RESTART. However, in practice, new cells are assigned very large minval offsets (e.g., 1_000_000_000_000), so this edge case is extremely unlikely.
I agree this edge case is unlikely.
"Sequence #{sequence.seq_name} already >= minval, skipping RESTART and MINVALUE"@fjedelhauser It seems clear to me that [Plan] Update application code to use new plan ... (#571422) is required to complete before this issue.
I agree that [Plan] Remove old plan associated columns and d... (#571424) may not be necessary.
But, please validate this along the way as you investigate, and implement each part. @bhrai and @jhyson are also good subject matter experts too.
Offshoot thread from https://gitlab.com/groups/gitlab-operating-model/-/work_items/901#note_3171437065
scalability
Even human authors do not have adequate information here to make good decisions around scalability. To pick a domain, database. We require database reviews to help here - and slowly added tooling like database-testing pipelines, and table_size constraints. Definitely lots to improve here.
Yes
What happens to current features when the https://gitlab.com/gitlab-com is in one non-isolated org, and https://gitlab.com/gitlab-org is in another non-isolated org ?
@niskhakova @peterhegman The standard should be that no URL logic exists in models.
Models are the wrong layer of responsiblity here. Models are primarily in charge of data access, and data persistence. Not display logic.
@jblack7 Small question - Is this epic going to be the delivery epic ? There's a few items from https://gitlab.com/gitlab-com/gl-infra/tenant-scale/organizations/team/-/work_items/28#work-items I would like individual issues to be created.
I strongly agree. Let's experiment. Don't build a big framework around a ~30 line MD file.