[Feature] OPS Alerts (#1119)

Author: ajanikow

CHANGELOG.md

@@ -1,12 +1,13 @@
 # Change Log
 
 ## [master](https://github.com/arangodb/kube-arangodb/tree/master) (N/A)
-- (Feature) Add new field to DeploymentReplicationStatus with details on DC2DC sync status
+- (Feature) Add new field to DeploymentReplicationStatus with details on DC2DC sync status=
 - (Feature) Early connections support
 - (Bugfix) Fix and document action timeouts
 - (Feature) Propagate sidecars' ports to a member's service
 - (Debug Package) Initial commit
 - (Feature) Detach PVC from deployment in Ordered indexing method
+- (Feature) OPS Alerts
 
 ## [1.2.16](https://github.com/arangodb/kube-arangodb/tree/1.2.16) (2022-09-14)
 - (Feature) Add ArangoDeployment ServerGroupStatus

docs/generated/metrics/README.md

@@ -14,6 +14,7 @@
 |                  [arangodb_operator_agency_cache_member_serving](./arangodb_operator_agency_cache_member_serving.md)                  | arangodb_operator | agency_cache |  Gauge  | Determines if agency member is reachable                                              |
 |                         [arangodb_operator_agency_cache_present](./arangodb_operator_agency_cache_present.md)                         | arangodb_operator | agency_cache |  Gauge  | Determines if local agency cache is present                                           |
 |                         [arangodb_operator_agency_cache_serving](./arangodb_operator_agency_cache_serving.md)                         | arangodb_operator | agency_cache |  Gauge  | Determines if agency is serving                                                       |
+|                            [arangodb_operator_engine_ops_alerts](./arangodb_operator_engine_ops_alerts.md)                            | arangodb_operator |    engine    | Counter | Counter for actions which requires ops attention                                      |
 |                      [arangodb_operator_engine_panics_recovered](./arangodb_operator_engine_panics_recovered.md)                      | arangodb_operator |    engine    | Counter | Number of Panics recovered inside Operator reconciliation loop                        |
 |      [arangodb_operator_members_unexpected_container_exit_codes](./arangodb_operator_members_unexpected_container_exit_codes.md)      | arangodb_operator |   members    | Counter | Counter of unexpected restarts in pod (Containers/InitContainers/EphemeralContainers) |
 |                           [arangodb_operator_rebalancer_enabled](./arangodb_operator_rebalancer_enabled.md)                           | arangodb_operator |  rebalancer  |  Gauge  | Determines if rebalancer is enabled                                                   |

docs/generated/metrics/arangodb_operator_engine_ops_alerts.md

@@ -0,0 +1,19 @@
+# arangodb_operator_engine_ops_alerts (Counter)
+
+## Description
+
+Counter for actions which requires ops attention
+
+## Labels
+
+|   Label   | Description          |
+|:---------:|:---------------------|
+| namespace | Deployment Namespace |
+|   name    | Deployment Name      |
+
+
+## Alerting
+
+| Priority |                       Query                        | Description                                 |
+|:--------:|:--------------------------------------------------:|:--------------------------------------------|
+| Warning  | irate(arangodb_operator_engine_ops_alerts[1m]) > 1 | Trigger an alert if OPS attention is needed |

internal/metrics.yaml

@@ -221,4 +221,18 @@ namespaces:
         labels:
           - key: section
             description: "Panic Section"
+      ops_alerts:
+        shortDescription: "Counter for actions which requires ops attention"
+        description: "Counter for actions which requires ops attention"
+        type: "Counter"
+        labels:
+          - key: namespace
+            description: "Deployment Namespace"
+          - key: name
+            description: "Deployment Name"
+        alertingRules:
+          - priority: Warning
+            query: irate(arangodb_operator_engine_ops_alerts[1m]) > 1
+            description: "Trigger an alert if OPS attention is needed"
+
 

pkg/deployment/context_impl.go

@@ -23,6 +23,7 @@ package deployment
 import (
 	"context"
 	"crypto/tls"
+	"fmt"
 	"net"
 	nhttp "net/http"
 	"strconv"
@@ -651,3 +652,13 @@ func (d *Deployment) GenerateMemberEndpoint(group api.ServerGroup, member api.Me
 func (d *Deployment) ACS() sutil.ACS {
 	return d.acs
 }
+
+func (d *Deployment) CreateOperatorEngineOpsAlertEvent(message string, args ...interface{}) {
+	if d == nil {
+		return
+	}
+
+	d.metrics.ArangodbOperatorEngineOpsAlerts++
+
+	d.CreateEvent(k8sutil.NewOperatorEngineOpsAlertEvent(fmt.Sprintf(message, args...), d.GetAPIObject()))
+}

pkg/deployment/metrics.go

@@ -36,6 +36,8 @@ type Metrics struct {
 		DeploymentValidationErrors, DeploymentImmutableErrors, StatusRestores uint64
 	}
 
+	ArangodbOperatorEngineOpsAlerts int
+
 	Deployment struct {
 		Accepted, UpToDate bool
 	}

pkg/deployment/reconcile/action_cleanout_member.go

@@ -141,6 +141,12 @@ func (a *actionCleanOutMember) CheckProgress(ctx context.Context) (bool, bool, e
 		}
 	}
 
+	if cache.PlanServers().Contains(agency.Server(m.ID)) {
+		// Something is wrong, servers is CleanedOut but still exists in the Plan
+		a.actionCtx.CreateOperatorEngineOpsAlertEvent("DBServer %s still exists in Plan after CleanOut", m.ID)
+		return false, true, nil
+	}
+
 	// Cleanout completed
 	return true, false, nil
 }

pkg/deployment/reconcile/action_context.go

@@ -51,6 +51,7 @@ type ActionContext interface {
 	reconciler.ArangoAgencyGet
 	reconciler.DeploymentInfoGetter
 	reconciler.DeploymentDatabaseClient
+	reconciler.KubernetesEventGenerator
 
 	member.StateInspectorGetter
 
@@ -130,6 +131,10 @@ type actionContext struct {
 	metrics      *Metrics
 }
 
+func (ac *actionContext) CreateOperatorEngineOpsAlertEvent(message string, args ...interface{}) {
+	ac.context.CreateOperatorEngineOpsAlertEvent(message, args...)
+}
+
 func (ac *actionContext) Metrics() *Metrics {
 	return ac.metrics
 }

pkg/deployment/reconcile/plan_builder_test.go

@@ -86,6 +86,11 @@ type testContext struct {
 	state     member.StateInspector
 }
 
+func (c *testContext) CreateOperatorEngineOpsAlertEvent(message string, args ...interface{}) {
+	//TODO implement me
+	panic("implement me")
+}
+
 func (c *testContext) GetAgencyHealth() (agencyCache.Health, bool) {
 	//TODO implement me
 	panic("implement me")

pkg/deployment/reconciler/context.go

@@ -138,6 +138,8 @@ type KubernetesEventGenerator interface {
 	// CreateEvent creates a given event.
 	// On error, the error is logged.
 	CreateEvent(evt *k8sutil.Event)
+
+	CreateOperatorEngineOpsAlertEvent(message string, args ...interface{})
 }
 
 // DeploymentClient provides functionalities to get deployment's clients.