Reduce CPU for Cloud SQL Auth Proxy

[priyadarshini-ni]

What does this Pull Request accomplish?

The CPU for Cloud SQL Auth Proxy has been reduced from 100m to 75m for the following service after aligning with owners

1. TestMonitor Service
2. Dynamic Form Fields

On performance testing with high mix, low volume scenario for TM, we observed the below utilization for 100m CPU and 100Mi memory requests

Query utilization:

Ingestion utilization:

In general, we noticed the CPU usage is around 15 to 30m and memory usage is around 20 to 30Mi in actuals when checked with kubectl top command.

As the pod autoscaling works based on the aggregate utilization of all the container resources, 100Mi (20% of TM service's RAM limit 512Mi and 30% of DFF's RAM limit 320Mi) is a bit small and shouldn't significantly affect autoscaling. Whereas 100m (40% of both service's CPU request 250m) CPU can be reduced a little till 75m as the utilization is less than 20% and also, we don't set limits for CPU. This will minimize any possibility of autoscaling issues.

Why should this Pull Request be merged?

The reduced CPU of 75m should be adequate as per the testing done.

What testing has been done?

Performance testing is done with 100m CPU and the utilization is around 20%, so reducing it to 75m shouldn't affect the performance.

Note

Dynamic Form Fields is not performance tested in GCP. Similar auth proxy resources as TM are added as default resources for DFF as

1. the service is less performance-intensive compared to TM
2. the service container's resource requests are similar to TM. So, autoscaling impact due to utilization of auth proxy resources should be similar