feat(redshift, rds): alarm on min/max connection count and long query duration (#251)

Adds alarm for min/max connection count and for long query duration.

Closes #248
Closes #249
Closes #232

---

_By submitting this pull request, I confirm that my contribution is made under the terms of the Apache-2.0 license_

Author: voho

API.md

@@ -92,8 +92,8 @@ You can browse the documentation at https://constructs.dev/packages/cdk-monitori
 | AWS Lambda (`.monitorLambdaFunction()`) | Latency, errors, iterator max age | Latency, errors, throttles, iterator max age | Optional Lambda Insights metrics (opt-in) support |
 | AWS Load Balancing (`.monitorNetworkLoadBalancer()`, `.monitorFargateApplicationLoadBalancer()`, `.monitorFargateNetworkLoadBalancer()`, `.monitorEc2ApplicationLoadBalancer()`, `.monitorEc2NetworkLoadBalancer()`) | System resources and task health | Unhealthy task count, running tasks count, (for Fargate/Ec2 apps) CPU/memory usage | Use for FargateService or Ec2Service backed by a NetworkLoadBalancer or ApplicationLoadBalancer |
 | AWS OpenSearch/Elasticsearch (`.monitorOpenSearchCluster()`, `.monitorElasticsearchCluster()`) | Indexing and search latency, disk/memory/CPU usage | Indexing and search latency, disk/memory/CPU usage, cluster status, KMS keys | |
-| AWS RDS (`.monitorRdsCluster()`) | Query duration, connections, latency, disk/CPU usage | Disk and CPU usage | |
-| AWS Redshift (`.monitorRedshiftCluster()`) | Query duration, connections, latency, disk/CPU usage | Disk and CPU usage | |
+| AWS RDS (`.monitorRdsCluster()`) | Query duration, connections, latency, disk/CPU usage | Connections, disk and CPU usage | |
+| AWS Redshift (`.monitorRedshiftCluster()`) | Query duration, connections, latency, disk/CPU usage | Query duration, connections, disk and CPU usage | |
 | AWS S3 Bucket (`.monitorS3Bucket()`) | Bucket size and number of objects | | |
 | AWS SecretsManager (`.monitorSecretsManagerSecret()`) | Days since last rotation | Days since last change or rotation | |
 | AWS SNS Topic (`.monitorSnsTopic()`) | Message count, size, failed notifications | Failed notifications, min/max published messages | |

README.md

@@ -0,0 +1,61 @@
+import {
+  ComparisonOperator,
+  TreatMissingData,
+} from "aws-cdk-lib/aws-cloudwatch";
+
+import { AlarmFactory, CustomAlarmThreshold } from "../../alarm";
+import { MetricWithAlarmSupport } from "../../metric";
+
+export interface LowConnectionCountThreshold extends CustomAlarmThreshold {
+  readonly minConnectionCount: number;
+}
+
+export interface HighConnectionCountThreshold extends CustomAlarmThreshold {
+  readonly maxConnectionCount: number;
+}
+
+export class ConnectionAlarmFactory {
+  protected readonly alarmFactory: AlarmFactory;
+
+  constructor(alarmFactory: AlarmFactory) {
+    this.alarmFactory = alarmFactory;
+  }
+
+  addMinConnectionCountAlarm(
+    metric: MetricWithAlarmSupport,
+    props: LowConnectionCountThreshold,
+    disambiguator?: string
+  ) {
+    return this.alarmFactory.addAlarm(metric, {
+      treatMissingData:
+        props.treatMissingDataOverride ?? TreatMissingData.MISSING,
+      comparisonOperator:
+        props.comparisonOperatorOverride ??
+        ComparisonOperator.LESS_THAN_THRESHOLD,
+      ...props,
+      disambiguator,
+      threshold: props.minConnectionCount,
+      alarmNameSuffix: "Connection-Count-Low",
+      alarmDescription: `Number of connections is too low.`,
+    });
+  }
+
+  addMaxConnectionCountAlarm(
+    metric: MetricWithAlarmSupport,
+    props: HighConnectionCountThreshold,
+    disambiguator?: string
+  ) {
+    return this.alarmFactory.addAlarm(metric, {
+      treatMissingData:
+        props.treatMissingDataOverride ?? TreatMissingData.MISSING,
+      comparisonOperator:
+        props.comparisonOperatorOverride ??
+        ComparisonOperator.GREATER_THAN_THRESHOLD,
+      ...props,
+      disambiguator,
+      threshold: props.maxConnectionCount,
+      alarmNameSuffix: "Connection-Count-High",
+      alarmDescription: `Number of connections is too high.`,
+    });
+  }
+}

lib/common/monitoring/alarms/ConnectionAlarmFactory.ts

@@ -1,6 +1,7 @@
 export * from "./AgeAlarmFactory";
 export * from "./AnomalyDetectingAlarmFactory";
 export * from "./CustomAlarmFactory";
+export * from "./ConnectionAlarmFactory";
 export * from "./DynamoAlarmFactory";
 export * from "./ElastiCacheAlarmFactory";
 export * from "./ErrorAlarmFactory";

lib/common/monitoring/alarms/index.ts

@@ -6,10 +6,13 @@ import {
 
 import {
   BaseMonitoringProps,
+  ConnectionAlarmFactory,
   CountAxisFromZero,
   DefaultGraphWidgetHeight,
   DefaultSummaryWidgetHeight,
   HalfWidth,
+  HighConnectionCountThreshold,
+  LowConnectionCountThreshold,
   MetricWithAlarmSupport,
   Monitoring,
   MonitoringScope,
@@ -32,6 +35,14 @@ import {
 export interface RdsClusterMonitoringOptions extends BaseMonitoringProps {
   readonly addDiskSpaceUsageAlarm?: Record<string, UsageThreshold>;
   readonly addCpuUsageAlarm?: Record<string, UsageThreshold>;
+  readonly addMinConnectionCountAlarm?: Record<
+    string,
+    LowConnectionCountThreshold
+  >;
+  readonly addMaxConnectionCountAlarm?: Record<
+    string,
+    HighConnectionCountThreshold
+  >;
 }
 
 export interface RdsClusterMonitoringProps
@@ -43,7 +54,9 @@ export class RdsClusterMonitoring extends Monitoring {
   readonly url?: string;
 
   readonly usageAlarmFactory: UsageAlarmFactory;
+  readonly connectionAlarmFactory: ConnectionAlarmFactory;
   readonly usageAnnotations: HorizontalAnnotation[];
+  readonly connectionAnnotations: HorizontalAnnotation[];
 
   readonly connectionsMetric: MetricWithAlarmSupport;
   readonly diskSpaceUsageMetric: MetricWithAlarmSupport;
@@ -83,7 +96,10 @@ export class RdsClusterMonitoring extends Monitoring {
       namingStrategy.resolveAlarmFriendlyName()
     );
     this.usageAlarmFactory = new UsageAlarmFactory(alarmFactory);
+    this.connectionAlarmFactory = new ConnectionAlarmFactory(alarmFactory);
+
     this.usageAnnotations = [];
+    this.connectionAnnotations = [];
 
     for (const disambiguator in props.addDiskSpaceUsageAlarm) {
       const alarmProps = props.addDiskSpaceUsageAlarm[disambiguator];
@@ -107,6 +123,30 @@ export class RdsClusterMonitoring extends Monitoring {
       this.addAlarm(createdAlarm);
     }
 
+    for (const disambiguator in props.addMinConnectionCountAlarm) {
+      const alarmProps = props.addMinConnectionCountAlarm[disambiguator];
+      const createdAlarm =
+        this.connectionAlarmFactory.addMinConnectionCountAlarm(
+          this.connectionsMetric,
+          alarmProps,
+          disambiguator
+        );
+      this.connectionAnnotations.push(createdAlarm.annotation);
+      this.addAlarm(createdAlarm);
+    }
+
+    for (const disambiguator in props.addMaxConnectionCountAlarm) {
+      const alarmProps = props.addMaxConnectionCountAlarm[disambiguator];
+      const createdAlarm =
+        this.connectionAlarmFactory.addMaxConnectionCountAlarm(
+          this.connectionsMetric,
+          alarmProps,
+          disambiguator
+        );
+      this.connectionAnnotations.push(createdAlarm.annotation);
+      this.addAlarm(createdAlarm);
+    }
+
     props.useCreatedAlarms?.consume(this.createdAlarms());
   }
 
@@ -154,6 +194,7 @@ export class RdsClusterMonitoring extends Monitoring {
       title: "Connections",
       left: [this.connectionsMetric],
       leftYAxis: CountAxisFromZero,
+      leftAnnotations: this.connectionAnnotations,
     });
   }
 

lib/monitoring/aws-rds/RdsClusterMonitoring.ts

@@ -8,9 +8,15 @@ import {
   BaseMonitoringProps,
   BooleanAxisFromZeroToOne,
   CountAxisFromZero,
+  ConnectionAlarmFactory,
   DefaultGraphWidgetHeight,
   DefaultSummaryWidgetHeight,
+  DurationThreshold,
   HalfQuarterWidth,
+  HighConnectionCountThreshold,
+  LatencyAlarmFactory,
+  LatencyType,
+  LowConnectionCountThreshold,
   MetricWithAlarmSupport,
   Monitoring,
   MonitoringScope,
@@ -33,6 +39,15 @@ import {
 export interface RedshiftClusterMonitoringOptions extends BaseMonitoringProps {
   readonly addDiskSpaceUsageAlarm?: Record<string, UsageThreshold>;
   readonly addCpuUsageAlarm?: Record<string, UsageThreshold>;
+  readonly addMaxLongQueryDurationAlarm?: Record<string, DurationThreshold>;
+  readonly addMinConnectionCountAlarm?: Record<
+    string,
+    LowConnectionCountThreshold
+  >;
+  readonly addMaxConnectionCountAlarm?: Record<
+    string,
+    HighConnectionCountThreshold
+  >;
 }
 
 export interface RedshiftClusterMonitoringProps
@@ -44,6 +59,10 @@ export class RedshiftClusterMonitoring extends Monitoring {
   readonly url?: string;
 
   readonly usageAlarmFactory: UsageAlarmFactory;
+  readonly latencyAlarmFactory: LatencyAlarmFactory;
+  readonly connectionAlarmFactory: ConnectionAlarmFactory;
+  readonly queryDurationAnnotations: HorizontalAnnotation[];
+  readonly connectionAnnotations: HorizontalAnnotation[];
   readonly usageAnnotations: HorizontalAnnotation[];
 
   readonly connectionsMetric: MetricWithAlarmSupport;
@@ -72,6 +91,10 @@ export class RedshiftClusterMonitoring extends Monitoring {
       namingStrategy.resolveAlarmFriendlyName()
     );
     this.usageAlarmFactory = new UsageAlarmFactory(alarmFactory);
+    this.latencyAlarmFactory = new LatencyAlarmFactory(alarmFactory);
+    this.connectionAlarmFactory = new ConnectionAlarmFactory(alarmFactory);
+    this.queryDurationAnnotations = [];
+    this.connectionAnnotations = [];
     this.usageAnnotations = [];
 
     const metricFactory = new RedshiftClusterMetricFactory(
@@ -114,6 +137,42 @@ export class RedshiftClusterMonitoring extends Monitoring {
       this.addAlarm(createdAlarm);
     }
 
+    for (const disambiguator in props.addMaxLongQueryDurationAlarm) {
+      const alarmProps = props.addMaxLongQueryDurationAlarm[disambiguator];
+      const createdAlarm = this.latencyAlarmFactory.addDurationAlarm(
+        this.longQueryDurationMetric,
+        LatencyType.P90,
+        alarmProps,
+        disambiguator
+      );
+      this.queryDurationAnnotations.push(createdAlarm.annotation);
+      this.addAlarm(createdAlarm);
+    }
+
+    for (const disambiguator in props.addMinConnectionCountAlarm) {
+      const alarmProps = props.addMinConnectionCountAlarm[disambiguator];
+      const createdAlarm =
+        this.connectionAlarmFactory.addMinConnectionCountAlarm(
+          this.connectionsMetric,
+          alarmProps,
+          disambiguator
+        );
+      this.connectionAnnotations.push(createdAlarm.annotation);
+      this.addAlarm(createdAlarm);
+    }
+
+    for (const disambiguator in props.addMaxConnectionCountAlarm) {
+      const alarmProps = props.addMaxConnectionCountAlarm[disambiguator];
+      const createdAlarm =
+        this.connectionAlarmFactory.addMaxConnectionCountAlarm(
+          this.connectionsMetric,
+          alarmProps,
+          disambiguator
+        );
+      this.connectionAnnotations.push(createdAlarm.annotation);
+      this.addAlarm(createdAlarm);
+    }
+
     props.useCreatedAlarms?.consume(this.createdAlarms());
   }
 
@@ -163,6 +222,7 @@ export class RedshiftClusterMonitoring extends Monitoring {
       title: "Connections",
       left: [this.connectionsMetric],
       leftYAxis: CountAxisFromZero,
+      leftAnnotations: this.connectionAnnotations,
     });
   }
 
@@ -177,6 +237,7 @@ export class RedshiftClusterMonitoring extends Monitoring {
         this.longQueryDurationMetric,
       ],
       leftYAxis: TimeAxisMillisFromZero,
+      leftAnnotations: this.queryDurationAnnotations,
     });
   }
 

lib/monitoring/aws-redshift/RedshiftClusterMonitoring.ts

@@ -38,6 +38,16 @@ test("snapshot test: all alarms", () => {
         maxUsagePercent: 70,
       },
     },
+    addMinConnectionCountAlarm: {
+      Warning: {
+        minConnectionCount: 1,
+      },
+    },
+    addMaxConnectionCountAlarm: {
+      Warning: {
+        maxConnectionCount: 100,
+      },
+    },
     useCreatedAlarms: {
       consume(alarms) {
         numAlarmsCreated = alarms.length;
@@ -46,6 +56,6 @@ test("snapshot test: all alarms", () => {
   });
 
   addMonitoringDashboardsToStack(stack, monitoring);
-  expect(numAlarmsCreated).toStrictEqual(2);
+  expect(numAlarmsCreated).toStrictEqual(4);
   expect(Template.fromStack(stack)).toMatchSnapshot();
 });