monitoring: add 2 new nvmeof alerts

Add NVMeoFMissingListener and NVMeoFZeroListenerSubsystem
alerts to prometheus_alerts.libsonnet.

Signed-off-by: Vallari Agrawal <vallari.agrawal@ibm.com>

diff --git a/monitoring/ceph-mixin/prometheus_alerts.libsonnet b/monitoring/ceph-mixin/prometheus_alerts.libsonnet
index cde1a736f8cd3de71fdb383ea8ba9cba5c7280e9..30b6b07d463bd30d529c988ef8684938a4eeb77e 100644 (file)
--- a/monitoring/ceph-mixin/prometheus_alerts.libsonnet
+++ b/monitoring/ceph-mixin/prometheus_alerts.libsonnet
@@ -935,6 +935,26 @@
             description: 'The supported limit for clients connecting to a subsystem is %(NVMeoFHighClientCount)d' % $._config,
           },
         },
+        {
+          alert: 'NVMeoFMissingListener',
+          'for': '10m',
+          expr: 'ceph_nvmeof_subsystem_listener_count == 0 and on(nqn) sum(ceph_nvmeof_subsystem_listener_count) by (nqn) > 0',
+          labels: { severity: 'warning', type: 'ceph_default' },
+          annotations: {
+            summary: 'No listener added for {{ $labels.instance }} NVMe-oF Gateway to {{ $labels.nqn }} subsystem',
+            description: 'For every subsystem, each gateway should have a listener to balance traffic between gateways.',
+          },
+        },
+        {
+          alert: 'NVMeoFZeroListenerSubsystem',
+          'for': '10m',
+          expr: 'sum(ceph_nvmeof_subsystem_listener_count) by (nqn) == 0',
+          labels: { severity: 'warning', type: 'ceph_default' },
+          annotations: {
+            summary: 'No listeners added to {{ $labels.nqn }} subsystem',
+            description: 'NVMeoF gateway configuration incomplete; one of the subsystems have zero listeners.',
+          },
+        },
         {
           alert: 'NVMeoFHighHostCPU',
           'for': '10m',