News for hackers. Stuff that matters.

ottovonbivouac

Tried with multiple phones and tablets both on Android and iOS. Did not manage to reproduce this problem. :S
The idea came when we tried to optimize Cluster filesystem (CEPH via Rook) for K8S. During this exercise we needed to test both PV&PVC performance, as well as TCP latency and bandwidths.

Target audience would be Kubernetes and OpenShift administrators who would like to measure and tune their installations. It is convenient for them as one CR can be used to describe the performance measurement.

You can use taints, tolerations, affinity and anti-affinity of k8s to schedule the load-generator to a different node than the one contains the service under test.

The load-gen and the service does not need to be co-located in case of REST (drill) and PostgreSQL (pgbench) based load testing.

In case of pure network benchmarking (with iperf & qperf) currently only the same cluster is supported.

Until then, could you please give the following page a try? https://kubestone.readthedocs.io/en/latest/

Thanks!

Thanks for the info, will try to investigate soon!
We did not plan it yet, but if there is interest we certainly can add it in the near future. We also welcome PRs :)

What we have as of today is Drill support, which is a HTTP load tester with Ansible like syntax:

- Drill GitHub page: https://github.com/fcsonline/drill

- kubestone doc for drill: https://kubestone.io/en/latest/benchmarks/drill/