Skip to content
This repository was archived by the owner on Jun 26, 2023. It is now read-only.

Latest commit

 

History

History
73 lines (58 loc) · 2.29 KB

README.md

File metadata and controls

73 lines (58 loc) · 2.29 KB

HNC Performance

Before you start

This performance test is to test HNC's startup performance, including controller working time and number of object reads and writes during HNC startup and restart.

Please be aware that the default performance test will create about maximum 500 namespaces and 1000 objects in your cluster. To change the scale, you can edit load-topology- files to update the namespace numbers by N and object numbers by O.

There are three topologies we support in this test to simulate possible use cases:

Wide - 1 root and N children (2-level-hierarchy)
Full - 1 root, N children and N*N grandchildren (3-level-hierarchy)
Skewer - 1 root, 1 child, 1 grandchild, 1 great grandchild, ... for N generations (N-level-hierarchy)

Run performance test

First, you will need to set the enable-test-log flag so that HNC can generate logs for this test to catch.

To do so, add - --enable-test-log in config/manager/manager.yaml so container args will look like this:

args:
  ...
- "--max-reconciles=10"
- "--apiserver-qps-throttle=50"
- "--enable-test-log"        ####### <<<<---- Add this line

Run make manifests to update the manifest to include the flag.

Now you can run performance test under the HNC root directory by:

$ scripts/performance/test.sh

You are expected to see a lot of namespace/role/rolebinding creation and eventually get a report like this:

HNC startup time : 2020-12-04T15:42:09
Current time : 2020-12-04T15:42:22
Controllers start working time : 2020-12-04T15:42:14
Controllers finish working time : 2020-12-04T15:42:19
Controllers working time for HNC startup: 5s
Total HierConfig reconciles: 33
Total Object reconciles: 217
Total HierConfig writes: 21
Total Namespace writes: 21
Total Object writes: 50

How to measure memory

To measure realtime memory usage, you can load one topology, deploy HNC and monitor the memory usage change manually. On GKE, you can find the realtime CPU and memory usages if you click the name of the workload in Workloads tab. It will show you the used memory compared with the limits and requests.

Here's an example:

$ scripts/performance/load-topology-full.sh
$ kubectl apply -f manifests/hnc-manager.yaml

### View memory usages now ###

# Cleanup namespaces
$ scripts/performance/clean-up-topologies.sh