Add fio-standalone Helm chart #1
Conversation
README.md
Outdated
| @@ -28,7 +28,7 @@ Latency in ns (Read/Write) | |||
| CPU Idleness: 72% | |||
| ``` | |||
|
|
|||
| ### Example Result of Comparsion Benchmark | |||
| ### Example Result of Comparison Benchmark | |||
There was a problem hiding this comment.
I've cleaned up the README.md and fix a few typos.
README.md
Outdated
| To deploy with all defaults: | ||
| ```bash | ||
| helm upgrade --install fio-benchmark-default \ | ||
| yasker-benchmark/fio-standalone \ |
There was a problem hiding this comment.
We can just call it fio. I think we should be able to cover both single and compare cases in the chart. Comparison is very useful, don't want to just leave it to yaml, lol.
There was a problem hiding this comment.
My thought was we could move a comparison benchmark into "install the helm chart multiple times" to allow N comparisons easily - maybe we could rework the README to have a YAML-based comparison and a Helm-based comparison? The Helm comparison would be something like helm install ... -f values-local.yaml; helm install ... -f values-longhorn.yaml.
| helm upgrade --install fio-benchmark-default \ | ||
| yasker-benchmark/fio-standalone \ | ||
| -n fio-benchmarks \ | ||
| --version v0.1.0 |
There was a problem hiding this comment.
Yeah, I should release a v0.1.0.
| -n fio-benchmarks \ | ||
| --version v0.1.0 \ | ||
| --set pvc.size=11Gi \ | ||
| --set benchmark.size=10G |
There was a problem hiding this comment.
Can we automatically set PVC size to benchmark size * 1.1? Not sure if the chart can do that though.
There was a problem hiding this comment.
Good call, will update templates and values
| @@ -0,0 +1,5 @@ | |||
| apiVersion: v2 | |||
| name: fio-standalone | |||
| @@ -0,0 +1,5 @@ | |||
| apiVersion: v2 | |||
| name: fio-standalone | |||
| description: Benchmark a single Kubernetes StorageClass | |||
There was a problem hiding this comment.
E.g. Benchmark and compare Kubernetes StorageClass performance.
| image: | ||
| repository: "yasker/benchmark" | ||
| pullPolicy: "IfNotPresent" | ||
| tag: "latest" |
README.md
Outdated
| @@ -92,9 +92,9 @@ Something is *wrong* when: | |||
| * For **Latency**, the **CPU Idleness** should be at least 40% to guarantee the test won't be impacted by CPU starving. | |||
| * If this happens, adding more CPUs to the node, or move to a beefer machine. | |||
|
|
|||
| ### Deploy the benchmark | |||
| ## Deploy the FIO benchmark | |||
There was a problem hiding this comment.
I am keeping it nested since I think we can add more benchmarks later.
Allow Helm-based installation for easier repeated deployments; add instructions and examples to README.