add some more prometheus rules for blackbox exporter

6 files changed, 80 insertions, 3 deletions

diff --git a/roles/monitoring/prometheus/server/defaults/main/main.yml b/roles/monitoring/prometheus/server/defaults/main/main.yml
index bae0cdba..09cd150c 100644
--- a/roles/monitoring/prometheus/server/defaults/main/main.yml
+++ b/roles/monitoring/prometheus/server/defaults/main/main.yml
@@ -14,7 +14,9 @@ prometheus_server_rules:
   prometheus: "{{ prometheus_server_rules_prometheus + ((prometheus_server_alertmanager is defined) | ternary(prometheus_server_rules_prometheus_alertmanager, [])) + prometheus_server_rules_prometheus_extra }}"
   node: "{{ prometheus_server_rules_node + prometheus_server_rules_node_extra }}"
   openwrt: "{{ prometheus_server_rules_openwrt + prometheus_server_rules_node_extra }}"
+  nut: "{{ prometheus_server_rules_nut + prometheus_server_rules_nut_extra }}"
   nut/ups: "{{ prometheus_server_rules_nut__ups + prometheus_server_rules_nut__ups_extra }}"
+  blackbox: "{{ prometheus_server_rules_blackbox + prometheus_server_rules_blackbox_extra }}"
   blackbox/ping: "{{ prometheus_server_rules_blackbox__ping + prometheus_server_rules_blackbox__ping_extra }}"
   blackbox/https: "{{ prometheus_server_rules_blackbox__https + prometheus_server_rules_blackbox__https_extra }}"
   blackbox/ssh: "{{ prometheus_server_rules_blackbox__ssh + prometheus_server_rules_blackbox__ssh_extra }}"
diff --git a/roles/monitoring/prometheus/server/defaults/main/rules_blackbox.yml b/roles/monitoring/prometheus/server/defaults/main/rules_blackbox.yml
new file mode 100644
index 00000000..d5c1fd42
--- /dev/null
+++ b/roles/monitoring/prometheus/server/defaults/main/rules_blackbox.yml
@@ -0,0 +1,47 @@
+---
+prometheus_server_rules_blackbox_extra: []
+prometheus_server_rules_blackbox:
+  - alert: BlackboxProbeFailed
+    expr: probe_success == 0
+    for: 0m
+    labels:
+      severity: critical
+    annotations:
+      summary: Blackbox probe failed (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "Probe failed\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
+
+  - alert: BlackboxSlowProbe
+    expr: avg_over_time(probe_duration_seconds[1m]) > 1
+    for: 1m
+    labels:
+      severity: warning
+    annotations:
+      summary: Blackbox slow probe (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "Blackbox probe took more than 1s to complete\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
+
+  - alert: BlackboxSslCertificateWillExpireSoon
+    expr: probe_ssl_earliest_cert_expiry - time() < 86400 * 30
+    for: 0m
+    labels:
+      severity: warning
+    annotations:
+      summary: Blackbox SSL certificate will expire soon (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "SSL certificate expires in 30 days\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
+
+  - alert: BlackboxSslCertificateWillExpireSoon
+    expr: probe_ssl_earliest_cert_expiry - time() < 86400 * 3
+    for: 0m
+    labels:
+      severity: critical
+    annotations:
+      summary: Blackbox SSL certificate will expire soon (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "SSL certificate expires in 3 days\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
+
+  - alert: BlackboxSslCertificateExpired
+    expr: probe_ssl_earliest_cert_expiry - time() <= 0
+    for: 0m
+    labels:
+      severity: critical
+    annotations:
+      summary: Blackbox SSL certificate expired (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "SSL certificate has expired already\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
diff --git a/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__https.yml b/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__https.yml
index cfdc10bd..140e3b4f 100644
--- a/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__https.yml
+++ b/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__https.yml
@@ -1,3 +1,20 @@
 ---
 prometheus_server_rules_blackbox__https_extra: []
-prometheus_server_rules_blackbox__https: []
+prometheus_server_rules_blackbox__https:
+  - alert: BlackboxProbeHttpFailure
+    expr: probe_http_status_code <= 199 OR probe_http_status_code >= 400
+    for: 0m
+    labels:
+      severity: critical
+    annotations:
+      summary: Blackbox probe HTTP failure (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "HTTP status code is not 200-399\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
+
+  - alert: BlackboxProbeSlowHttp
+    expr: avg_over_time(probe_http_duration_seconds[1m]) > 1
+    for: 1m
+    labels:
+      severity: warning
+    annotations:
+      summary: Blackbox probe slow HTTP (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "HTTP request took more than 1s\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
diff --git a/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__ping.yml b/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__ping.yml
index 06ce8607..cc87b6b1 100644
--- a/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__ping.yml
+++ b/roles/monitoring/prometheus/server/defaults/main/rules_blackbox__ping.yml
@@ -1,3 +1,11 @@
 ---
 prometheus_server_rules_blackbox__ping_extra: []
-prometheus_server_rules_blackbox__ping: []
+prometheus_server_rules_blackbox__ping:
+  - alert: BlackboxProbeSlowPing
+    expr: avg_over_time(probe_icmp_duration_seconds[1m]) > 1
+    for: 1m
+    labels:
+      severity: warning
+    annotations:
+      summary: Blackbox probe slow ping (instance {{ '{{' }} $labels.instance {{ '}}' }})
+      description: "Blackbox ping took more than 1s\n  VALUE = {{ '{{' }} $value {{ '}}' }}\n  LABELS = {{ '{{' }} $labels {{ '}}' }}"
diff --git a/roles/monitoring/prometheus/server/defaults/main/rules_nut.yml b/roles/monitoring/prometheus/server/defaults/main/rules_nut.yml
new file mode 100644
index 00000000..d8d64f64
--- /dev/null
+++ b/roles/monitoring/prometheus/server/defaults/main/rules_nut.yml
@@ -0,0 +1,3 @@
+---
+prometheus_server_rules_nut_extra: []
+prometheus_server_rules_nut: []
diff --git a/roles/monitoring/prometheus/server/tasks/main.yml b/roles/monitoring/prometheus/server/tasks/main.yml
index 16167c9c..c0928cc3 100644
--- a/roles/monitoring/prometheus/server/tasks/main.yml
+++ b/roles/monitoring/prometheus/server/tasks/main.yml
@@ -83,7 +83,7 @@
     state: directory
 
 - name: generate rules files for all jobs
-  loop: "{{ prometheus_server_jobs | union(['prometheus']) }}"
+  loop: "{{ prometheus_server_jobs | union(['prometheus']) | union(prometheus_server_jobs | select('match', '.*/.*') | map('dirname') | unique) }}"
   template:
     src: rules.yml.j2
     dest: "/etc/prometheus/rules/{{ item }}.yml"