kesamba opened a new pull request, #9216:
URL: https://github.com/apache/devlake/pull/9216
### Summary
When several DevLake projects point at the **same Jira board** and each
narrows it with its own ExtraJQL, the DORA metrics that depend on incidents
(Change Failure Rate, MTTR / failed deployment recovery time) counted the
**whole board** for every project, so one project's incidents leaked into the
others' numbers.
This PR records which project collected each board issue and filters on it
end to end. It is split into commits that each build and pass tests on their
own, so it can be reviewed commit by commit:
1. **`feat(core): add project_name to board_issues`**: new `project_name`
column, part of the primary key so an issue in two projects' slices can be
linked once per project, with default `''`. Migration `20260820`. Existing e2e
snapshots that verify `board_issues` gain the (empty) column.
2. **`feat(jira): record the collecting project on board issues`**:
`project_name` on `_tool_jira_board_issues` (migration `20260710`) and
`ProjectName` on `JiraApiParams`. The issue collector and extractor stamp the
pipeline's project. `ProjectName` is `omitempty` in the serialized raw-data
params, so pipelines without a project keep exactly the same params (no
re-collection and no orphaned raw data on upgrade).
3. **`feat(jira): convert only the current project's slice of a shared
board`**: the issue convertor reads only the current project's tool rows and
stamps the project on domain `board_issues`.
4. **`fix(jira): resync the board when its ExtraJQL changes`**: the rendered
ExtraJQL is folded into the board filter's change-detection key, so changing
ExtraJQL triggers a full resync and issues that no longer match get unlinked
from the board.
5. **`feat(dora): scope incidents and deployment links to the project's
slice`**: `ConvertIssuesToIncidents` and `ConnectIncidentToDeployment` filter
on `board_issues.project_name`.
6. **`feat(grafana): filter DORA incident panels by project`**: the same
filter in the incident queries of the DORA dashboards (mysql: DORA,
dora-by-team, dora-debug, dora-details-change-failure-rate,
dora-details-failed-deployment-recovery-time,
dora-details-timeto-restore-service; postgresql: DORA, dora-by-team).
**Backward compatibility:** if a board has *no* per-project stamps at all
(single-project setups, or data collected before this change), every query
keeps the previous whole-board behavior. If a board *does* have stamps,
unstamped or NULL rows are treated as an ambiguous leftover pool and excluded,
instead of being attributed to every project.
### Does this close any open issues?
N/A
### Screenshots
N/A
### Other Information
- New unit tests run the real production queries against an in-memory SQLite
DB. This adds `gorm.io/driver/sqlite` as a direct dependency; it was already in
`go.sum`.
- `plugins/jira/tasks/issue_convertor_project_filter_test.go`
- `plugins/jira/tasks/issue_extractor_test.go`
(`TestExtractIssuesStampsProjectName`)
- `plugins/dora/tasks/incident_from_issue_generator_test.go`
- `plugins/dora/tasks/incident_deploy_connector_test.go`
- The dora queries quote the reserved `table` column per dialect (backticks
on MySQL, double quotes otherwise), so they run on MySQL, Postgres and SQLite.
- The full upstream CI passed on the fork, including `test-e2e` on both
MySQL and Postgres (all plugin e2e suites plus `test/e2e/remote` and
`test/e2e/services`), golangci-lint, migration-script-lint and
check-grafana-dashboards: https://github.com/kesamba/devlake/pull/1
🤖 Generated with [Claude Code](https://claude.com/claude-code)
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
