Merge pull request newrelic#2302 from pkudikyala/NR-229976

Temporal new Quickstart.

alert-policies/temporal/ActivityEexcutionFail.yml

@@ -0,0 +1,31 @@
+# Name of the alert
+name: Activity Execution Fails
+
+# Description and details
+description: |+
+  This alert is triggered if the Temporal activity execution fails once within a 5 minutes window.
+# Type of alert
+type: STATIC
+
+# NRQL query
+nrql:
+
+  query: "SELECT latest(temporal_activity_execution_failed_total) AS 'execution failed' 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: 1
+    # 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/temporal/ActivityExecutionLatency.yml

@@ -0,0 +1,40 @@
+# Name of the alert
+name: Activity Excution Latency
+
+# Description and details
+description: |+
+  This alert is triggered if the Temporal activity execution latency exceeds 5 seconds for 5 minutes.
+# Type of alert
+type: STATIC
+
+# NRQL query
+nrql:
+
+  query: "SELECT sum(temporal_activity_execution_latency_seconds_sum) 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: 5
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 300
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+  - priority: WARNING
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 4
+    # Time in seconds; 120 - 3600, must be a multiple of 60 for Baseline conditions
+    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/temporal/ActivityScheduleStartLatency.yml

@@ -0,0 +1,40 @@
+# Name of the alert
+name: Activity Schedule to Start Latency
+
+# Description and details
+description: |+
+  This alert is triggered if the Temporal activity schedule to start latency exceeds 5 seconds for 5 minutes.
+# Type of alert
+type: STATIC
+
+# NRQL query
+nrql:
+
+  query: "SELECT sum(temporal_activity_schedule_to_start_latency_sum) 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: 5
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 300
+    # How many data points must be in violation for the duration
+    thresholdOccurrences: ALL
+  - priority: WARNING
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 4
+    # Time in seconds; 120 - 3600, must be a multiple of 60 for Baseline conditions
+    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/temporal/MemoryUsagePercent.yml

@@ -0,0 +1,40 @@
+# Name of the alert
+name: Memory Usage Percent
+
+# Description and details
+description: |+
+  This alert is triggered if the memory usage exceeds 90% for 5 minutes.
+# Type of alert
+type: STATIC
+
+# NRQL query
+nrql:
+
+  query: "SELECT average(memoryUsedPercent) AS `Memory used %` FROM SystemSample"
+
+# 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
+  - priority: WARNING
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 85
+    # Time in seconds; 120 - 3600, must be a multiple of 60 for Baseline conditions
+    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