Init runner sub command

[fuweid]

This pull request is to implement kperf runner run sub-command.

How to build

# cd into repo
$ make build
$ ls -al ./bin
total 40404
drwxrwxr-x 2 weifu weifu     4096 Dec  8 07:16 .
drwxrwxr-x 9 weifu weifu     4096 Dec  8 02:55 ..
-rwxrwxr-x 1 weifu weifu 41364238 Dec  8 07:16 kperf

$ ./bin/kperf runner run -h
NAME:
   kperf runner run - run a benchmark test to kube-apiserver

USAGE:
   kperf runner run [command options] [arguments...]

OPTIONS:
   --kubeconfig value  Path to the kubeconfig file
   --config value      Path to the configuration file
   --conns value       Total number of connections. It can override corresponding value defined by --config (default: 1)
   --rate value        Maximum requests per second (Zero means no limitation). It can override corresponding value defined by --config (default: 0)
   --total value       Total number of requests. It can override corresponding value defined by --config (default: 1000)
   --user-agent value  User Agent

How to install

The user can install the binary into /usr/local/bin by default.

$ make install

The user can change the path by PREFIX.

# install binaries into /tmp/bin
$ PREFIX=/tmp make install

How to build binary into image

Default image name is localhost:5000/kperf:latest.

$ make image-build

The users can use IMAGE_REPO and IMAGE_TAG to change the image name.
For instance,

$ IMAGE_REPO=examplee2e.azurecr.io/public IMAGE_TAG=v1 make image-build
building examplee2e.azurecr.io/public/kperf:v1
....

How to push image

Both IMAGE_REPO and IMAGE_TAG also apply to image-push recipe.

$ make image-push

How to run benchmark test?

• Step 1: Prepare load test profile

The profile is defined at api/types/load_profile.go. There is an example.

$ cat /tmp/xx.yaml
version: 1
description: test
spec:
  rate: 100
  total: 10
  conns: 10
  requests:
  - staleList:
      version: v1
      resource: pods
      limit: 500
    shares: 100
  - quorumList:
      version: v1
      resource: pods
      limit: 1000
    shares: 150

• Step 2: Prepare cluster

• Step 3: Run that load test with kubeconfig.

$ bin/kperf runner run --kubeconfig /home/weifu/.kube/config --config /tmp/xx.yaml --user-agent testing --total=10
Response stat:
  Total: 10
  Failures: 0
  Duration: 1.93187487s
  Requests/sec: 5.18
  Latency Distribution:
    [0.00] 0.138s
    [0.50] 0.141s
    [0.90] 0.144s
    [0.95] 0.663s
    [0.99] 0.663s
    [1.00] 0.663s

The kperf runner run will use service account if there is no kubeconfig.
If you use it in pod, please ensure that service account has permission to access.

[fuweid]

Merged it since CI is green.