import 15.2.2 octopus source

[ceph.git] / ceph / monitoring / prometheus / alerts / ceph_default_alerts.yml

diff --git a/ceph/monitoring/prometheus/alerts/ceph_default_alerts.yml b/ceph/monitoring/prometheus/alerts/ceph_default_alerts.yml
index a9b0560d4730b8ff9736d110c9377f39e25f44ec..aebd816156066196da0b4eec3374ceab36d0fe89 100644 (file)
--- a/ceph/monitoring/prometheus/alerts/ceph_default_alerts.yml
+++ b/ceph/monitoring/prometheus/alerts/ceph_default_alerts.yml
@@ -166,6 +166,7 @@ groups:
     rules:
       - alert: root volume full
         expr: node_filesystem_avail_bytes{mountpoint="/"} / node_filesystem_size_bytes{mountpoint="/"} * 100 < 5
+        for: 5m
         labels:
           severity: critical
           type: ceph_default
@@ -199,15 +200,10 @@ groups:
             Node {{ $labels.instance }} experiences packet errors > 1
             packet/s on interface {{ $labels.device }}.
 
-      # predict fs fill-up times
-      - alert: storage filling
+      - alert: storage filling up
         expr: |
-          (
-            (
-              node_filesystem_free_bytes / deriv(node_filesystem_free_bytes[2d])
-              * on(instance) group_left(nodename) node_uname_info
-            ) <= 5
-          ) > 0
+          predict_linear(node_filesystem_free_bytes[2d], 3600 * 24 * 5) *
+          on(instance) group_left(nodename) node_uname_info < 0
         labels:
           severity: warning
           type: ceph_default
@@ -234,10 +230,9 @@ groups:
       - alert: pool filling up
         expr: |
           (
-            (
-              (ceph_pool_max_avail - ceph_pool_stored) / deriv(ceph_pool_max_avail[2d])
-            ) * on(pool_id) group_right ceph_pool_metadata <= 5
-          ) > 0
+            predict_linear(ceph_pool_stored[2d], 3600 * 24 * 5) >=
+            ceph_pool_max_avail
+          ) * on(pool_id) group_right(name) ceph_pool_metadata
         labels:
           severity: warning
           type: ceph_default