Temporal Cloud: Updated the dashboards, alerts and config files.

alert-policies/temporal-cloud/FailedWorkflows.yml

@@ -3,12 +3,12 @@ name: Failed Workflows
 
 # Description and details
 description: |+
-  This alert is triggered if the Temporal cloud workflows fail once within a 5-minute window.
+  This alert is triggered if the Temporal Cloud workflows fail once within a 5-minute window.
 type: STATIC
 
 # NRQL query
 nrql:
-  query: "FROM temporalCloudWorkflowFailed SELECT latest(`data.result-value1`) FACET `data.result-metric-__name__`"
+  query: "FROM temporalCloudWorkflowFailed SELECT max(`data.result-value1`)AS 'Failed Workflows' FACET `data.result-metric-operation` AS operation, `data.result-metric-temporal_namespace` AS namespace"
 
 # Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
 valueFunction: SINGLE_VALUE

alert-policies/temporal-cloud/RPSLimitErrors.yml

@@ -0,0 +1,30 @@
+# Name of the alert
+name: Resource Exhausted Errors
+
+# Description and details
+description: |+
+  This alert is triggered if the Temporal Cloud resource exhausted errors exceeds 1 for 5 minutes.
+# Type of alert
+type: STATIC
+
+# NRQL query
+nrql:
+  query: "SELECT max(`data.result-value1`) AS resource_exhausted_errors FROM temporalCloudResourceExhaustedErrors FACET `data.result-metric-operation` AS operation, `data.result-metric-temporal_namespace` AS namespace"
+
+# 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-cloud/ServiceLatency.yml

@@ -1,15 +1,15 @@
 # Name of the alert
-name: Service Latency
+name: Start Workflow Execution Latency
 
 # Description and details
 description: |+
-  This alert is triggered if the Temporal cloud service latency exceeds 5 seconds for 5 minutes.
+  This alert is triggered if the Temporal Cloud service latency exceeds 5 seconds for 5 minutes.
 # Type of alert
 type: STATIC
 
 # NRQL query
 nrql:
-  query: "FROM temporalCloudWorkflowFailed SELECT latest(`data.result-value1`) FACET `data.result-metric-__name__`"
+  query: "SELECT max(`data.result-value1`) FROM temporalCloudStartWorkflowExecutionLatencyP95 FACET `data.result-metric-operation` AS operation, `data.result-metric-temporal_namespace` AS namespace"
 
 # Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE)
 valueFunction: SINGLE_VALUE
@@ -20,7 +20,7 @@ terms:
     # Operator used to compare against the threshold.
     operator: ABOVE
     # Value that triggers a violation
-    threshold: 1
+    threshold: 5
     # Time in seconds; 120 - 3600
     thresholdDuration: 300
     # How many data points must be in violation for the duration