Updated the dashboard and alerts.

alert-policies/nextcloud/ScrapeErrors.yml → alert-policies/nextcloud/CPUUsage.yml

@@ -1,11 +1,11 @@
-name: Scrape Errors
+name: CPU Usage
 
 description: |+
-  This alert is triggered  when the cache size limit is reached above 4 for at least 5 minutes.
+  This alert is triggered  when the CPU Usage is exceeds 90% for at least 5 minutes.
 
 type: STATIC
 nrql:
-  query: "FROM Metric SELECT sum(nextcloud_scrape_errors_total) As 'Scrape Errors'"
+  query: "SELECT average(host.cpuPercent) AS '(%) used CPU' FROM Metric"
 
 # Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE).
 valueFunction: SINGLE_VALUE
@@ -16,7 +16,7 @@ terms:
     # Operator used to compare against the threshold.
     operator: ABOVE
     # Value that triggers a violation.
-    threshold: 4  
+    threshold: 90 
     # Time in seconds; 120 - 3600
     thresholdDuration: 300
     # How many data points must be in violation for the duration?
@@ -26,7 +26,7 @@ terms:
     # Operator used to compare against the threshold.
     operator: ABOVE
     # Value that triggers a violation.
-    threshold: 1  
+    threshold: 80 
     # Time in seconds; 120 - 3600
     thresholdDuration: 300
     # How many data points must be in violation for the duration?

alert-policies/nextcloud/HeapMemoryUsage.yml → alert-policies/nextcloud/MemoryUsage.yml

@@ -1,11 +1,11 @@
-name: Heap Memory Usage
+name: Memory Usage 
 
 description: |+
-  This alert is triggered  when the Heap Memory Usage is below 2 for at least 5 minutes.
+  This alert is triggered  when the Memory Usage is exceeds 90% for at least 5 minutes.
 
 type: STATIC
 nrql:
-  query: "FROM Metric select latest(go_memstats_heap_inuse_bytes)/1e+6 As 'Heap Memory Usage'"
+  query: "FROM NextcloudFlexMetrics select (latest(ocs.data.server.php.opcache.memory_usage.used_memory) / 1e+6) / (latest(ocs.data.server.php.opcache.memory_usage.free_memory) / 1e+6) * 100 AS '(%) used memory'"
 
 # Function used to aggregate the NRQL query value(s) for comparison to the terms.threshold (Default: SINGLE_VALUE).
 valueFunction: SINGLE_VALUE
@@ -14,19 +14,19 @@ valueFunction: SINGLE_VALUE
 terms:
   - priority: CRITICAL
     # Operator used to compare against the threshold.
-    operator: BELOW
+    operator: ABOVE
     # Value that triggers a violation.
-    threshold: 2  
+    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: BELOW
+    operator: ABOVE
     # Value that triggers a violation.
-    threshold: 4  
+    threshold: 80 
     # Time in seconds; 120 - 3600
     thresholdDuration: 300
     # How many data points must be in violation for the duration?