* run annotation data migration in batches
* how could i miss it
* run in the background, starting from newest annotations
* update tests
* optionally pass batch size via env
* Suggestions: Deprecate previous API, enable external plugin suggestions behind flag
* fix types for deprecated builder
* restore some support for cloud-onboarding
* add support for cloud-onboarding usage, add test to ensure it keeps working
* refactor to not hardcode on 'core:'
* remove unused import
Change "Service Center" to "Service center" in navigation menu to follow
sentence case capitalization style consistently across the application.
Fixesgrafana/slo#3818🤖 Generated with [Claude Code](https://claude.com/claude-code)
Co-authored-by: Claude <noreply@anthropic.com>
We have two historians in alerting - alert state and notification. The intention
of this app is to provide query capabilities for both.
In this initial commit, the existing /history API is simply cloned to the new
app. It is identical except that it will send Kubernetes-style error responses
instead of Grafana-style.
This approach was taken to implement the new app more iteratively - ideally we
would define a new API, but this requires quite a significant overhaul of the
backend code.
* SchemaV2: Convertion from v1beta1 to v2beta1
* Compare backend-frontend v1 convertion
* Compare backend-frontend v1 convertion
* Fix fe be diff
* Resolve DS issues
* Fix ds inconsistecnies
* fix legacy string value issues
* fix ds test
* fix layout issue
* update test
* Fix tests and issue with defaultConfig
* Update output
* Fix viz config convertion
* wip
* Fix v1 to v2 dashboard transformation differences
Major fixes implemented:
- Backend function names in conversion.go
- Backend group field logic for queries, annotations, and vizConfig
- Backend datasource resolution with map-based lookup
- Backend timezone handling (empty string vs browser)
- Backend annotation processing (empty array vs default annotation)
- Backend default values (editable, liveNow)
- Backend variable processing (definition, defaultKeys, refresh, refId)
- Backend panel layout (y position calculations)
- Backend VizConfig (Kind and Group fields, default values)
- Frontend snapshot issue (annotations not processing)
- Frontend datasource references (only when original has valid datasource)
Test results:
- annotation-conversions: PASSING (0 differences)
- dashboard-properties: 3 expected architectural differences
- panel-conversions: Multiple expected architectural differences
- variable-conversions: 7 expected architectural differences
All remaining differences are expected architectural choices between
backend persistence optimization and frontend UI consumption optimization.
* fix issues with panel and annotation queries with no datasource
* definition and regex
* Use proper v1beta1 resource when testing
* remove misc file
* fix ds provider test
* fix def ds test in response transformer
* fix remaining ResponseTransformers test
* timesettings, variable refresh, editable, liveNow, definition
* fix transformSceneToSaveModelSchemaV2 test
* revert legacyRow changes
* fix go lint issues
* normalize y coordinates when serializing a row
* clean up
* update tests
* use GetStringValue from schemaversion
* fix go lint - cyclomatic complexity
* update open api snapshot
* add migrated dashboards
* fix default panel type when panel type is not provided
* revert dash link changes for now
* fix
* fix nested panel issue and default ref in v1
* apply defaults to nested panels too
* update snapshots
* fix issues with annotations
* matchers, showLegend, annotations
* when converting also don't process queries that have only a refId
* fix issues with text var
* fix dash links
* default to collapse: false when serializing
* fix: filter refId from variable query specs in backend migration
- Add buildDataQueryKindForVariable function to filter refId for variables
- Remove default refId "A" in transformSingleQuery
- Only include __legacyStringValue for non-empty string queries
- Remove refId addition in transformSaveModelSchemaV2ToScene.getDataQueryForVariable
- Handle undefined queries gracefully in frontend and backend
- Ensure backend matches frontend behavior for query variable serialization
* fix: default variable refresh to 'never' to match frontend behavior
Change backend default for missing refresh field from 'onDashboardLoad'
to 'never' to match frontend defaultVariableRefresh() schema default
* fix: only include iconColor in annotations when it exists
- Frontend: Use defaultAnnotationQuerySpec().iconColor as fallback to match schema defaults
- Backend: Only set iconColor if it exists in v1 input (not using GetStringValue)
- Ensures iconColor is only included when present in original dashboard
* fix: use schema defaults for annotation enable, hide, and iconColor
- Use defaultAnnotationQuerySpec() to get schema defaults instead of hardcoded values
- Default enable to false (schema default) to match frontend behavior
- Use schema default for iconColor and hide fields
- Ensures consistency with frontend which uses defaultAnnotationQuerySpec() defaults
* fix: set collapse for hidden-header rows to match first explicit row
- When panels appear before the first explicit row, the hidden-header row's
collapse should match the first explicit row's collapsed value
- Matches frontend behavior where collapse: panel.collapsed uses the next
row panel's collapsed value
- Ensures consistency between frontend and backend when converting rows layout
* fix: handle constant variables with missing query value
- Frontend: Fix bug where undefined value was converted to string 'undefined'
- Now defaults to empty string when value is undefined: value ? String(value) : ''
- Backend: Match frontend fix - default to empty string for text/value when query is missing
- Ensures consistency when constant variable query is missing from v1 dashboard
* Fix interval variable handling when query is missing
- Extract intervals from options when query is missing/empty (matches backend behavior)
- Handle undefined/null query in getIntervalsFromQueryString
- Handle missing current object/value in getCurrentValueForOldIntervalModel
- Update interval variable refresh to use literal 'onTimeRangeChanged' in schema
- Use defaultIntervalVariableSpec() for interval variable serialization
- Backend: Generate query string from options when query is missing
* Fix corrupted dashboard with systemRef override
* don't resolve types for template variables in datasource refs on the backend
* fix annotation and ds issues
* fix range and special mappings
* fix datasource var pluginId and regex
* add __systemRef to schema
* update v15 migration annotation to have a ds type because v2 keeps track of if type is in the initial save model, and if it's not it removes it, but for frontendOuput we are running transformSaveModelToScene which will then assign the type
* add migration fields since the backend applies automigrations in collapsed rows
* filter out queries in ResponseTransformer that only have refId field
* lint
* v2: add default query if queries are empty to match v1 behavior
* fix single migration test
* tracking test should have a defined spec otherwise datasource is removed and won't be tracked
* initialize default with default ds ref
* wip
* Do not assign DS if ds group is empty
* cleanup
* revert change in setupTests.ts
* clean up TODO
* query with only refId should not expect to have a group
* refactor: extract v0alpha1 to v1beta1 conversion logic into atomic function
- Extract ConvertDashboard_V0_to_V1beta1 into v0alpha1_to_v1beta1.go
- Extract prepareV0ConversionContext and migrateV0Dashboard helper functions
- Standardize v0.go to match v1.go pattern with inline multi-step conversions
- Implement Convert_V0_to_V2alpha1 using atomic functions (v0->v1beta1->v2alpha1)
- Implement Convert_V0_to_V2beta1 using atomic functions (v0->v1beta1->v2alpha1->v2beta1)
- Remove non-atomic v0alpha1_to_v2alpha1.go file
* test: add version-specific test files for conversion error handling
- Extract v0 conversion tests into v0_test.go
- Extract v1 conversion tests into v1_test.go
- Add v2 conversion tests in v2_test.go
- Ensure all error handling paths in conversion functions are covered
- Add tests for Convert_V0_to_V2alpha1 and Convert_V0_to_V2beta1 error paths
- Add tests for Convert_V1beta1_to_V2alpha1 and Convert_V1beta1_to_V2beta1 error paths
- Add tests for Convert_V2alpha1_to_V2beta1 error handling
* Fix tests
* Fix linter
* Clean up
* feat(dashboard): Add automatic data loss detection for dashboard conversions
Implements comprehensive data loss detection for all dashboard API version conversions.
Components Tracked:
• Panels (visualization + library panels)
• Queries (data source queries, excludes row panel queries)
• Annotations
• Links
• Variables (template variables)
Features:
• Automatic detection via withConversionMetrics wrapper (zero code changes)
• Error type: 'conversion_data_loss_error'
• Logs: panelsLost, queriesLost, annotationsLost, linksLost, variablesLost
Bugs Found:
• Fixed critical bug: metrics.go was silently swallowing ALL errors (return nil → return err)
Testing:
• TestDataLossDetectionOnAllInputFiles - runs all conversions with detailed logging
• V2→V0/V1 downgrades write output for debugging then skip (not yet implemented)
• All tests passing
* Run dashboards on schema v2 E2Es
* reveret unintended changes
* cleanup
* Reset active manager correctly according to toggles config
* Fix new dashboard being serialized as v1
* Rename toggle
---------
Co-authored-by: Ivan Ortega <ivanortegaalba@gmail.com>
Co-authored-by: Dominik Prokop <dominik.prokop@grafana.com>
* provisioning: acquire server lock before provisioning dashboards+folders
Signed-off-by: Maicon Costa <maiconscosta@gmail.com>
---------
Signed-off-by: Maicon Costa <maiconscosta@gmail.com>
Fix search for existing provisioned dashboards in modes 3+
The search query was not requesting the dashboard's "legacy ID". As a result,
the provisioning process would not find existing provisioned dashboards, making
copies of these dashboards every time there was a change in the provisioned
dashboard's definition.
Part 2 of adding version messages to the `alert_rule_version` able. This allows
setting the message via a header when using the Prometheus conversion API,
which can be useful for e.g. linking changes back to source control.
* fix: collate rule_group column as utf8mb4_bin
This ensures this column is sorted identically by the database and
golang, which should eliminate issues related to a mismatch in sort
order.
* chore: make test stricter
* fix: also add binary collation for postgres
To better observe and identify performance bottlenecks in the prometheus rules API.
The following spans were added:
- `api.prometheus.RouteGetRuleStatuses`
- `api.prometheus.PrepareRuleGroupStatusesV2`
The `api.prometheus.PrepareRuleGroupStatusesV2` includes attributes to capture the parameters used in the request to help with debugging and performance analysis.
* Deprecating features.IsEnabled
* add one more nolint
* add one more nolint
* Give better hints to devs in the deprecation message of IsEnabledGlobally
* adding more doc strings
* fix linter after rebase
* Extend deprecation message
