Screwdriver: Aggregate view for a pipeline

[jithine]

What happened:

Users do not have a view to see the overall status of a pipeline.

What you expected to happen:

UI Mock

1. Provide an aggregate button in UI to display a workflow graph which shows latest build across current active jobs.
2. API should support an end point which returns latest build across current active jobs.

How to reproduce it:

N/A

[jithine]

Few more pending items

• [x] Remove link from underlink of build bubbles.
• [x] Build order list should be oldest -> newst. Reverse current order.
• [ ] Actions Play button should start a fresh event and not restart last successful
• [x] Data for coverage column
• [ ] Job order should be based on position in workflow graph.
  1. Group Jobs based on depth in workflow graph
  2. Within a job group sort alphabetically
• [ ] Build stats context window - should display SCM sha and an option to restart

[tk3fftk]

It looks great feature 👍
By the way, is it OK that models has specific queries? Should these implementation are inside of datastore implementation?
Also, should we have database specific queries? It's may be hard to know where should be updated when we or someone add another database implementation.

[adong]

It looks great feature 👍
By the way, is it OK that models has specific queries? Should these implementation are inside of datastore implementation?
Also, should we have database specific queries? It's may be hard to know where should be updated when we or someone add another database implementation.

Are you referring to PR: https://github.com/screwdriver-cd/models/pull/441/files#diff-43e1dd855c9ebdae88cbf0102c6e05b2R8

Good question, let me escalate

[tk3fftk]

@adong Thank you for escalation. Should I create another issue?

[jithine]

@tk3fftk My replies inline

By the way, is it OK that models has specific queries? Should these implementation are inside of datastore implementation?

I believe reasoning was we don't have datastore specific to DB engines like MySQL, only sequalize level with datastore-sequalize and complete raw queries need to be passed over to the datastore-sequalize with all parameters substituted. These raw queries contains data which is specific to model business logic, like the DB fields like jobIDs and values in them, we don't have any representation for those in the datastore implementation. And our datastore expects complete data to be passed over to them, so to enforce that separation of business logic specific data, raw queries was put in models

Also, should we have database specific queries? It's may be hard to know where should be updated when we or someone add another database implementation.

Challenge is having a common query which works for all DB implementations. Based on the use case we had, we were not able to find a single query which works in all DB implementations. If we can find one then we should get rid of current separate queries. And sequalize was of no help in doing these queries in a generic way. The generic problem was "Select top n matches for DB rows and group them together" eg: latest n builds for all the job ids which are sorted by created time

[tk3fftk]

I got it. these seem to be needed. But I think we should avoid frequent use of that.

[jithine]

Re-opening to address UX alignment with mocks.