Merge pull request #37 from anjackson/master

Minor monitoring updates
ukwa · Jan 11, 2022 · 69d7e88 · 69d7e88
2 parents 4dfbac8 + f864167
commit 69d7e88
Show file tree

Hide file tree

Showing 2 changed files with 10 additions and 9 deletions.
diff --git a/monitor/prometheus/alert.rules.yml b/monitor/prometheus/alert.rules.yml
@@ -37,14 +37,15 @@ groups:
       summary: "Task {{ $labels.job }} failed to run"
       description: "Job {{ $labels.job }} failed to run successfully."
 
-  - alert: nominet_task_has_not_run
-    expr: absent(ukwa_task_event_timestamp{job="NominetDomainListToHDFS"}) or (time() - ukwa_task_event_timestamp{job="NominetDomainListToHDFS", status="event.core.success"} ) / (60*60*24) > 31
-    for: 2h
-    labels:
-      severity: severe
-    annotations:
-      summary: "Task {{ $labels.job }} failed to run"
-      description: "Job {{ $labels.job }} failed to run successfully."
+# Commenting this out until we have an agreement in place:
+#  - alert: nominet_task_has_not_run
+#    expr: absent(ukwa_task_event_timestamp{job="NominetDomainListToHDFS"}) or (time() - ukwa_task_event_timestamp{job="NominetDomainListToHDFS", status="event.core.success"} ) / (60*60*24) > 31
+#    for: 2h
+#    labels:
+#      severity: severe
+#    annotations:
+#      summary: "Task {{ $labels.job }} failed to run"
+#      description: "Job {{ $labels.job }} failed to run successfully."
 
   - alert: crawl_launcher_has_not_run
     expr: absent(ukwa_task_event_timestamp{job="crawl.LaunchCrawls"}) or (time() - ukwa_task_event_timestamp{job='crawl.LaunchCrawls', status="event.core.success"}) > 3600

diff --git a/monitor/prometheus/prometheus.yml-template b/monitor/prometheus/prometheus.yml-template
@@ -167,7 +167,7 @@ scrape_configs:
     metrics_path: /admin/metrics/
     static_configs:
       - targets: 
-        - 'ingest:5050'
+        - 'prod1:5050'
 
   - job_name: 'solr'
     static_configs: