use NoSchedule taint data in system information in metrics report rather than reporting it repeatedly

clearlinux / cloud-native-setup

Automation around setting up the cloud-native content (kubernetes) on Clear Linux.

Apache License 2.0

61 stars 41 forks source link

use NoSchedule taint data in system information in metrics report rather than reporting it repeatedly #219

Open dklyle opened 5 years ago

dklyle commented 5 years ago

The system config data in the json files already contains the NoSchedule taint information. This information does not change during a test run. Use the system config information rather than repeatedly reporting the NoSchedule taint per node, per iteration. This will avoid unnecessary overhead of making execs to the stats pods.

dklyle commented 5 years ago

I'm actually inclined to leave this as is for the k8s_scale.sh, it's fixed in k8s_scale_rapid.sh, but the node information is used in the k8s_scale.sh as well. While it would be possible to extract it. I'm not sure the overall benefit is worth it at this point.