This commit fixes a bug where we did not send resolved alerts to Alertmanager for resolved alert instances. This meant that resolved notifications did not have the annotations from the resolved state, and a result did not also have the resolved screenshot.
Prior to this change, all alert instance writes and deletes happened
individually, in their own database transaction. This change batches up
writes or deletes for a given rule's evaluation loop into a single
transaction before applying it.
Before:
```
goos: darwin
goarch: arm64
pkg: github.com/grafana/grafana/pkg/services/ngalert/store
BenchmarkAlertInstanceOperations-8 398 2991381 ns/op 1133537 B/op 27703 allocs/op
--- BENCH: BenchmarkAlertInstanceOperations-8
util.go:127: alert definition: {orgID: 1, UID: FovKXiRVzm} with title: "an alert definition FTvFXmRVkz" interval: 60 created
util.go:127: alert definition: {orgID: 1, UID: foDFXmRVkm} with title: "an alert definition fovFXmRVkz" interval: 60 created
util.go:127: alert definition: {orgID: 1, UID: VQvFuigVkm} with title: "an alert definition VwDKXmR4kz" interval: 60 created
PASS
ok github.com/grafana/grafana/pkg/services/ngalert/store 1.619s
```
After:
```
goos: darwin
goarch: arm64
pkg: github.com/grafana/grafana/pkg/services/ngalert/store
BenchmarkAlertInstanceOperations-8 1440 816484 ns/op 352297 B/op 6529 allocs/op
--- BENCH: BenchmarkAlertInstanceOperations-8
util.go:127: alert definition: {orgID: 1, UID: 302r_igVzm} with title: "an alert definition q0h9lmR4zz" interval: 60 created
util.go:127: alert definition: {orgID: 1, UID: 71hrlmR4km} with title: "an alert definition nJ29_mR4zz" interval: 60 created
util.go:127: alert definition: {orgID: 1, UID: Cahr_mR4zm} with title: "an alert definition ja2rlmg4zz" interval: 60 created
PASS
ok github.com/grafana/grafana/pkg/services/ngalert/store 1.383s
```
So we cut time by about 75% and memory allocations by about 60% when
storing and deleting 100 instances.
This change also updates some of our tests so that they run successfully against postgreSQL - we were using random Int64s, but postgres integers, which our tables use, max out at 2^31-1
* move saving the state to state manager when scheduler stops
* move saving state to ProcessEvalResults
* add GetRuleKey to State
* add LogContext to AlertRuleKey
* add tests for cache getOrCreate
* update ProcessEvalResults to accept extra lables
* extract to getRuleExtraLabels
* move populating of constant rule labels to extra labels
This commit fixes a bug where the state did not change from Alerting to Error if the evaluation result returned an error, or from Error to Alerting if evaluations stopped returning errors.
* Introduce AlertsRouter in the sender package, and move all fields and methods related to notifications out of the scheduler to this router.
* Introduce a new interface AlertsSender in the schedule package and replace calls of anonymous function `notify` inside the ruleRoutine to calling methods of that interface.
* Rename interface Scheduler in api package to ExternalAlertmanagerProvider, and replace scheduler with AlertRouter as struct that implements the interface.
* Alerting: decapitalize log lines and use "err" as the key for errors
Found using (logger|log).(Warn|Debug|Info|Error)\([A-Z] and (logger|log).(Warn|Debug|Info|Error)\(.+"error"
Adds three functions:
`withStoredImages` iterates over a list of models.Alerts, extracting a stored image's data from storage, if available, and executing a user-provided function.
`withStoredImage` does this for an image attached to a specific alert.
`openImage` finds and opens an image file on disk.
Moves `store.Image` to `models.Image`
Simplifies `channels.ImageStore` interface and updates notifiers that use it to use the simpler methods.
Updates all pkg/alert/notifier/channels to use withStoredImage routines.
This change adds a field to state.State and models.AlertInstance
that indicate the "Reason" that an instance has its current state. This
helps us account for cases where the state is "Normal" but the
underlying evaluation returned "NoData" or "Error", for example.
Fixes#42606
Signed-off-by: Joe Blubaugh <joe.blubaugh@grafana.com>
The State Manager will now take screenshots when an alert instance
switches to an Alerting or Resolved state.
Signed-off-by: Joe Blubaugh joe.blubaugh@grafana.com
This commit adds a pkg/services/screenshot package for taking and uploading screenshots of Grafana dashboards. It supports taking screenshots of both dashboards and individual panels within a dashboard, using the rendering service.
The screenshot package has the following services, most of which can be composed:
BrowserScreenshotService (Takes screenshots with headless Chrome)
CachableScreenshotService (Caches screenshots taken with another service such as BrowserScreenshotService)
NoopScreenshotService (A no-op screenshot service for tests)
SingleFlightScreenshotService (Prevents duplicate screenshots when taking screenshots of the same dashboard or panel in parallel)
ScreenshotUnavailableService (A screenshot service that returns ErrScreenshotsUnavailable)
UploadingScreenshotService (A screenshot service that uploads taken screenshots)
The screenshot package does not support wire dependency injection yet. ngalert constructs its own version of the service. See https://github.com/grafana/grafana/issues/49296
This PR also adds an ImageScreenshotService to ngAlert. This is used to take screenshots with a screenshotservice and then store their location reference for use by alert instances and notifiers.
* rename folder to match package name
* backend/sqlstore: move GetDashboard into DashboardService
This is a stepping-stone commit which copies the GetDashboard function - which lets us remove the sqlstore from the interfaces in dashboards - without changing any other callers.
* checkpoint: moving GetDashboard calls into dashboard service
* finish refactoring api tests for dashboardService.GetDashboard
* Alerting: Accurately set value for prom-compatible APIs
Sets the value fields for the prometheus compatible API based on a combination of condition `refID` and the values extracted from the different frames.
* Fix an extra test
* Ensure a consitent ordering
* Address review comments
* address review comments
* Alerting: Remove internal labels from prometheus compatible API responses
* Appease the linter
* Fix integration tests
* Fix API documentation & linter
* move removal of internal labels to the models
This commit changes staleResultsHandler to create an annotation if the current state is Alerting and the result is being removed from the state cache as it has not been updated since 2x the evaluation interval.
* Fix Annotation creation
- Remove validation of panelID, now annotations are created irrespective on whether they're attached to a panel or not.
- Alwasy attach the annotation to an AlertID
* Fix annotation creation
* fix tests
* Alerting: Clear alerting rule evaluation errors after intermittent failures
When an alert transitioned in a way that `alerting -> error -> (alerting|nodata)`, the error provided by the `error` state would never be cleared thus the API and UI would show the health as an error.
* do not suppress NoData state
* extract conversion of state to postable alert + tests
* create a special alert instance if nodata
* use NoData when converting from Keep Last State instead of Alerting
* add silence during migration if NoData is mapped to KeepLastState.
Yuriy Tseretyan
Sofia Papagiannaki
George Robinson
Joe Blubaugh
gotjosh
Kristin Laemmert
idafurjes
Yuriy Tseretyan
gotjosh
Santiago