Open ashutoshrathore opened 7 months ago
@aimichelle FYI
@ashutoshrathore gave this a try once #1887 was merged. While this fixed the issues with the DaemonSet pods, the expected behaviors was to be able to control the scheduling of the OLM and operator pods as well. #1913 implements this by adding two new node selector configuration options in the helm chart.
Is your feature request related to a problem? Please describe. Pixie deploys various pods like vizier-query broker, vizier cloud connector, kelvin and vizier-pem.
When we try to add nodeSelector for these pods using values defined here , it also selected the node pool for vizier-pem and that actually disable the function of vizier-pem to run on all nodes in cluster.
Pem already has pre-defined value for memory i.e. pemMemoryRequest and pemMemoryLimit
Describe the solution you'd like nodeSelector should only work for pods that dont require to run on all nodes like query broker or kelvin. Vizier-pem should have separate resource blocks. Rather if we can have separate defination for each deployments then it will help to customise in a better way for users.
Describe alternatives you've considered No alternatives available
Additional context NA