Skip to content

monitoring: wrap cadvisor metrics in label_replace

Warren Gifford requested to merge distribution/provisioning-alerts-fix into master

Created by: bobheadxi

Closes https://github.com/sourcegraph/sourcegraph/issues/11571

It seems the name label has some special undocumented purpose, which was causing the provisioning alert_rules to fail to create.

This PR wraps each of the cAdvisor dashboards with a label_replace to add a container label that is just the value of the name label. I manually updated Prometheus on k8s.sgdev.org and it seems to do the trick for alert_counts that were previously failing:

Screen Shot 2020-07-02 at 5 25 37 PM

This PR also includes two misc. improvements:

  • remove (not available on server) from individual metrics - this text already shows up on dashboard section titles
  • fix formatting for possible solutions

Merge request reports

Loading