kubernetes
/
community
mirror of https://github.com/kubernetes/community.git
1
0
Fork 0
Code Issues Packages Projects Releases Wiki Activity

conventional test args to bypass UTs 

- when running benchmark tests, we used to use `-run=xxx` to exclude unit tests; however a more conventional way is to use `-run=^$`, `-run=^$$` here is to excape the $ simbol, otherwise it won't work.
- add info to display benchmark output only
This commit is contained in:
Wei Huang 2019-06-19 14:26:12 +08:00
parent 0a406008ed
commit 7ce57c6a8e
No known key found for this signature in database
GPG Key ID: BE5E9752F8B6E005
1 changed files with 6 additions and 4 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
contributors/devel/sig-scheduling/scheduler_benchmarking.md
View File

@ -11,7 +11,7 @@ To run integration benchmarks use the following command from inside a Kubernetes
directory.
```sh
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=xxx -bench=."
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -bench=."
```
You can also provide a benchmark name in order to run a specific set of
@ -19,9 +19,11 @@ benchmarks. Please refer to [Go documentation on benchmarks](https://golang.org/
for more information.
```sh
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=xxx -bench=BenchmarkScheduling"
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -bench=BenchmarkScheduling"
```
> To display benchmark output only, you can append `-alsologtostderr=false -logtostderr=false` to `KUBE_TEST_ARGS`.
These benchmarks are located in `./test/integration/scheduler_perf/scheduler_bench_test.go`.
The function names start with `BenchmarkScheduling`. At the beginning of each
function there are a few lines in the form of:
@ -51,7 +53,7 @@ it in the `-bench` argument. For example, the following will run only those
benchmarks with 5000 nodes and 1000 existing pods.
```sh
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=xxx -bench=BenchmarkScheduling/5000Nodes/1000Pods"
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -bench=BenchmarkScheduling/5000Nodes/1000Pods"
```
## Profiling the scheduler
@ -60,7 +62,7 @@ You can get CPU profiling information from the benchmarks by adding a `-cpuprofi
to the command above. Example:
```sh
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=xxx -bench=BenchmarkScheduling -cpuprofile cpu.out"
make test-integration WHAT=./test/integration/scheduler_perf KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -bench=BenchmarkScheduling -cpuprofile cpu.out"
```
After obtaining the CPU profile, you can use `pprof` to view the results. For