Merge branch 'main' into NR-89203

alert-policies/adobe-commerce-business-insights/5xxErrors.yml

@@ -0,0 +1,35 @@
+name: 5xx Server Errors
+
+description: |+
+  This alert is triggered if the customer faces 5xx server errors more than 5 times in 5 minutes.
+type: STATIC
+nrql:
+  query: "SELECT count(*) as '5xx Server Errors' from Transaction WHERE httpResponseCode LIKE '5%'"
+
+# 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: 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: 5
+    # 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/adobe-commerce-business-insights/cpuUsage.yml

@@ -0,0 +1,35 @@
+name: CPU Usage (%)
+
+description: |+
+  This alert is triggered if CPU usage exceeds 90% for 5 minutes.
+type: STATIC
+nrql:
+  query: "SELECT latest(host.cpuPercent) AS 'CPU Used %' 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
+  - priority: WARNING
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 80
+    # 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/adobe-commerce-business-insights/downtime.yml

@@ -0,0 +1,35 @@
+name: Downtime (%)
+
+description: |+
+  This alert is triggered if Downtime is more than 1% for 2 minutes.
+type: STATIC
+nrql:
+  query: "SELECT percentage(count(result), where result = 'FAILED') as 'Downtime (%)' from SyntheticCheck"
+
+# 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: 120
+    # 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: 0.5
+    # Time in seconds; 120 - 3600
+    thresholdDuration: 120
+    # 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/adobe-commerce-business-insights/memoryUsage.yml

@@ -0,0 +1,35 @@
+name: Memory Usage (%)
+
+description: |+
+  This alert is triggered if Memory usage exceeds 90% for 5 minutes.
+type: STATIC
+nrql:
+  query: "SELECT latest(host.memoryUsedPercent) as 'Memory Used %' 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
+  - priority: WARNING
+    # Operator used to compare against the threshold.
+    operator: ABOVE
+    # Value that triggers a violation
+    threshold: 80
+    # 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/vertica/DiskUsed.yml

@@ -0,0 +1,37 @@
+name: Disk Space Used (%)
+
+description: |+
+  This alert is triggered when disk space usage exceeds 90% for 5 minutes.
+type: STATIC
+nrql:
+  query: "SELECT latest(disk_space_free_percent) as 'Disk Space Used (%)' FROM VerticaDiskStorage"
+
+# 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
+    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
+