github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2026-07-19 07:25:19 +00:00
Code Issues Projects Releases Wiki Activity

scheduler_perf: document benchstat usage

Browse Source 
With benchstat it's easy to do before/after comparisons, but the section for
running benchmark didn't mention it at all and didn't work as shown there:
- benchmark results must be printed (FULL_LOG)
- timeout might have been too short (KUBE_TIMEOUT)
- only "short" benchmarks ran (SHORT)
- klog log output must be redirected (ARTIFACTS)
This commit is contained in:
Patrick Ohly
2025-06-26 07:57:59 +02:00
parent 84cacae704
commit 7910b43cd3
1 changed files with 16 additions and 4 deletions
Download Patch File Download Diff File Expand all files Collapse all files

20
test/integration/scheduler_perf/README.md
View File

@@ -33,9 +33,14 @@ Currently the test suite has the following:
```shell
# In Kubernetes root path
make test-integration WHAT=./test/integration/scheduler_perf/... ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -benchtime=1ns -bench=BenchmarkPerfScheduling"
make test-integration WHAT=./test/integration/scheduler_perf/... KUBE_TIMEOUT=-timeout=1h ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" FULL_LOG=y ARTIFACTS=/tmp SHORT=-short=false KUBE_TEST_ARGS='-run=^$ -benchtime=1ns -bench=BenchmarkPerfScheduling'
```
The output can used for [`benchstat`](https://pkg.go.dev/golang.org/x/perf/cmd/benchstat)
to summarize results or to do before/after
comparisons. Add `-count=6` to `KUBE_TEST_ARGS` to get statistically relevant
summaries.
The benchmark suite runs all the tests specified under subdirectories split by topic (`<topic>/performance-config.yaml`).
By default, it runs all workloads that have the "performance" label. In the configuration,
labels can be added to a test case and/or individual workloads. Each workload also has
@@ -46,7 +51,7 @@ a comma-separated list of label names. Each label may have a `+` or `-` as prefi
be set. For example, this runs all performance benchmarks except those that are labeled
as "integration-test":
```shell
make test-integration WHAT=./test/integration/scheduler_perf/... ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -benchtime=1ns -bench=BenchmarkPerfScheduling -perf-scheduling-label-filter=performance,-integration-test"
make test-integration WHAT=./test/integration/scheduler_perf/... KUBE_TIMEOUT=-timeout=1h ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" SHORT=-short=false ARTIFACTS=/tmp FULL_LOG=y KUBE_TEST_ARGS='-run=^$ -benchtime=1ns -bench=BenchmarkPerfScheduling -perf-scheduling-label-filter=performance,-integration-test'
```
Once the benchmark is finished, JSON files with metrics are available in the subdirectories (`test/integration/scheduler_perf/config/<topic>`).
@@ -60,16 +65,23 @@ Otherwise, the golang benchmark framework will try to run a test more than once
```shell
# In Kubernetes root path
make test-integration WHAT=./test/integration/scheduler_perf/... ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -benchtime=1ns -bench=BenchmarkPerfScheduling/SchedulingBasic/5000Nodes/5000InitPods/1000PodsToSchedule"
make test-integration WHAT=./test/integration/scheduler_perf/... KUBE_TIMEOUT=-timeout=1h ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" SHORT=-short=false ARTIFACTS=/tmp FULL_LOG=y KUBE_TEST_ARGS='-run=^$ -benchtime=1ns -bench=BenchmarkPerfScheduling/SchedulingBasic/5000Nodes/5000InitPods/1000PodsToSchedule'
```
To produce a cpu profile:
```shell
# In Kubernetes root path
make test-integration WHAT=./test/integration/scheduler_perf/... KUBE_TIMEOUT="-timeout=3600s" ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" KUBE_TEST_ARGS="-run=^$$ -benchtime=1ns -bench=BenchmarkPerfScheduling -cpuprofile ~/cpu-profile.out"
make test-integration WHAT=./test/integration/scheduler_perf/... KUBE_TIMEOUT=-timeout=1h ETCD_LOGLEVEL=warn KUBE_TEST_VMODULE="''" FULL_LOG=y ARTIFACTS=/tmp SHORT=-short=false KUBE_TEST_ARGS='-run=^$ -benchtime=1ns -bench=BenchmarkPerfScheduling -cpuprofile ~/cpu-profile.out'
```
Here some explanations for those parameters:
- `KUBE_TIMEOUT=-timeout=1h`: benchmarks may have to run longer than the default 10 minutes
- `FULL_LOG=y`: ensures that benchmark results are shown
- `ARTIFACTS=/tmp`: redirects non-test log output, which is lengthy and breaks parsing by `benchstat`
- `SHORT=-short=false`: enables running benchmarks which run sufficiently long to be meaningful
- `-run=^$`: disables running unit tests
### How to configure benchmark tests
Configuration files located under `<topic>/performance-config.yaml` contain a list of templates.