Skip to main content

Monitoring improvements

· 2 min read
Thibaut Raballand
Creator of Spider

Since Spider moved to the Kube, I needed to fix some of the monitoring dashboards to work on Kubernetes.

Here are the changes:

Elasticsearch metrics​

Elasticsearch cluster is now hosted on the Kube an not on specific servers.

The monitoring UI now supports both situation: it detects if it is hosted on Kube or not, and adapts the queries in consequence.

Main dashboard changes​

The main dashboard got a few changes:

  • ES disk Gaugues is monitoring ES persistent volumes
  • ES CPU and RAM Gaugues have their maximum set to the resource limits of ES PODs
  • Nodes status are showing all Nodes of the Kubernetes cluster, Spider being on them or not.
  • A Redis status box has been added to monitor Redis CPU and RAM usage, taking into account the maximum RAM set in setup.

MainDashboard.png

Logs dashboard changes​

I improved logs dashboard since the two main charts (warning and errors) where not very useful:

  • Logs by service is now showing how many errors and warning logs have been generated by service
  • Logs by Whisperer is now showing how many where linked to what Whisperer
    • If all logs are from same Whisperer, the issue is more on the observed environment than on Spider

Logs.png

Whisperer dashboard changes​

The Whisperers upload to server graph is now showing the total of uploaded data by whisperer, not by instance.
This is more relevant.

Whisperers.png