I have observed high latency on E2 standard nodes which correlates with high Scheduler Wait Time when using E2 nodes as part of a GKE node pool.
I was expecting Scheduler Wait Time to be much lower than observed and for latency not to be impacted to this extent.
This documentation seems to suggest that I should expect scheduler wait time to be very low. https://cloud.google.com/blog/products/compute/understanding-dynamic-resource-management-in-e2-vms
I have attached an example graph of the scheduler wait time and the corresponding APDEX for a service running on the node over an afternoon this week.
I'd be interested to know if there is anything I can do to avoid this impact on latency whilst still using E2 nodes.
The E2 series contain shared-core machine types. These machine types timeshare a physical core which can be a cost-effective method for running small, non-resource intensive apps.
If the E2 machine type cannot handle your workloads, the N2 series offers balanced price/performance across a wide range of machine types. You can refer to this table which provides recommendations for different workloads.
I am using standard E2 nodes, which as I understand it are not shared-core machine types.