Turn honor_timestamps off for resource scrape config

Metrics for pods were included in queries for some period after the pod was already removed. This resulted in inaccurate values for autoscaling. Turning honor_timestamps off for the kubernetes-resource-metrics job seems to fix this issue. Signed-off-by: Han Verstraete (OpenFaaS Ltd) <[email protected]>
openfaas · Aug 23, 2024 · 8664296 · 8664296
1 parent 180c3ab
commit 8664296
Showing 1 changed file with 1 addition and 0 deletions.
diff --git a/chart/openfaas/templates/prometheus-pro-cfg.yaml b/chart/openfaas/templates/prometheus-pro-cfg.yaml
@@ -98,6 +98,7 @@ data:
 
       - job_name: 'kubernetes-resource-metrics'
         scrape_interval: 10s
+        honor_timestamps: false
         scheme: https
         tls_config:
           ca_file: /var/run/secrets/kubernetes.io/serviceaccount/ca.crt