Merge branch 'main' into NR-164846

alert-policies/amazon-keyspaces/HighAccountProvisionedReadCapacityUtilization.yml

@@ -0,0 +1,27 @@
+name: High Account Provisioned Read Capacity Utilization
+
+description: |+
+  This alert is triggered when the Account Provisioned Read Capacity Utilization is above 90%.
+
+type: STATIC
+nrql:
+  query: "SELECT average(`aws.cassandra.AccountProvisionedReadCapacityUtilization`) FROM Metric"
+
+# Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
+valueFunction: SINGLE_VALUE
+
+# List of Critical and Warning thresholds for the condition
+terms:
+  - priority: CRITICAL
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 90
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 300
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+
+# Duration after which a violation automatically closes
+# Time in seconds; 300 - 2592000 (Default: 86400 [1 day])
+violationTimeLimitSeconds: 86400

alert-policies/amazon-keyspaces/HighAccountProvisionedWriteCapacityUtilization.yml

@@ -0,0 +1,27 @@
+name: High Account Provisioned Write Capacity Utilization
+
+description: |+
+  This alert is triggered when the Account Provisioned Write Capacity Utilization is above 90%.
+
+type: STATIC
+nrql:
+  query: "SELECT average(`aws.cassandra.AccountProvisionedWriteCapacityUtilization`) FROM Metric"
+
+# Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
+valueFunction: SINGLE_VALUE
+
+# List of Critical and Warning thresholds for the condition
+terms:
+  - priority: CRITICAL
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 90
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 300
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+
+# Duration after which a violation automatically closes
+# Time in seconds; 300 - 2592000 (Default: 86400 [1 day])
+violationTimeLimitSeconds: 86400

alert-policies/amazon-keyspaces/HighSystemErrors.yml

@@ -0,0 +1,27 @@
+name: High System Errors
+
+description: |+
+  This alert is triggered when System Errors are above 10 in 10 minutes.
+
+type: STATIC
+nrql:
+  query: "SELECT count(`aws.cassandra.SystemErrors`) FROM Metric"
+
+# Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
+valueFunction: SINGLE_VALUE
+
+# List of Critical and Warning thresholds for the condition
+terms:
+  - priority: CRITICAL
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 10
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 600
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+
+# Duration after which a violation automatically closes
+# Time in seconds; 300 - 2592000 (Default: 86400 [1 day])
+violationTimeLimitSeconds: 86400

alert-policies/amazon-keyspaces/HighUserErrors.yml

@@ -0,0 +1,27 @@
+name: High User Errors
+
+description: |+
+  This alert is triggered when User Errors are above 10 in 10 minutes.
+
+type: STATIC
+nrql:
+  query: "SELECT count(`aws.cassandra.UserErrors`) FROM Metric"
+
+# Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
+valueFunction: SINGLE_VALUE
+
+# List of Critical and Warning thresholds for the condition
+terms:
+  - priority: CRITICAL
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 10
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 600
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+
+# Duration after which a violation automatically closes
+# Time in seconds; 300 - 2592000 (Default: 86400 [1 day])
+violationTimeLimitSeconds: 86400

alert-policies/amazon-transfer-family/InboundFailedMessage.yml

@@ -0,0 +1,34 @@
+name: High Inbound Failed Message
+
+description: |+
+  This alert is triggered if the number of Inbound Failed Messages exceeds 20 for 10 minutes.
+
+type: STATIC
+nrql:
+  query: "SELECT sum(`aws.transfer.InboundFailedMessage`) as 'Query' FROM Metric"
+
+# Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
+valueFunction: SINGLE_VALUE
+
+# List of Critical and Warning thresholds for the condition
+terms:
+  - priority: CRITICAL
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 20
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 600
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+
+  # Adding a Warning threshold is optional
+  - priority: WARNING
+    operator: ABOVE
+    threshold: 10
+    thresholdDuration: 600
+    thresholdOccurrences: ALL
+
+# Duration after which a violation automatically closes
+# Time in seconds; 300 - 2592000 (Default: 86400 [1 day])
+violationTimeLimitSeconds: 86400

alert-policies/amazon-transfer-family/OnPartialUploadExecutionsFailed.yml

@@ -0,0 +1,34 @@
+name: High Partial Upload Executions Failed
+
+description: |+
+  This alert is triggered if the number of Partial Upload Execution Failure exceeds 20 for 10 minutes.
+
+type: STATIC
+nrql:
+  query: "SELECT sum(`aws.transfer.OnPartialUploadExecutionsFailed`) as 'Query' FROM Metric"
+
+# Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
+valueFunction: SINGLE_VALUE
+
+# List of Critical and Warning thresholds for the condition
+terms:
+  - priority: CRITICAL
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 20
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 600
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+
+  # Adding a Warning threshold is optional
+  - priority: WARNING
+    operator: ABOVE
+    threshold: 10
+    thresholdDuration: 600
+    thresholdOccurrences: ALL
+
+# Duration after which a violation automatically closes
+# Time in seconds; 300 - 2592000 (Default: 86400 [1 day])
+violationTimeLimitSeconds: 86400

alert-policies/amazon-transfer-family/OnUploadExecutionsFailed.yml

@@ -0,0 +1,34 @@
+name: High Upload Executions Failed
+
+description: |+
+  This alert is triggered if the number of Upload Execution Failure exceeds 20 for 10 minutes.
+
+type: STATIC
+nrql:
+  query: "SELECT sum(`aws.transfer.OnUploadExecutionsFailed`) as 'Query' FROM Metric"
+
+# Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
+valueFunction: SINGLE_VALUE
+
+# List of Critical and Warning thresholds for the condition
+terms:
+  - priority: CRITICAL
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 20
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 600
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+
+  # Adding a Warning threshold is optional
+  - priority: WARNING
+    operator: ABOVE
+    threshold: 10
+    thresholdDuration: 600
+    thresholdOccurrences: ALL
+
+# Duration after which a violation automatically closes
+# Time in seconds; 300 - 2592000 (Default: 86400 [1 day])
+violationTimeLimitSeconds: 86400