Robert Krawitz
Robert Krawitz
To better understand: If I specify `connections/sec` of 5, I'd open a connection, send traffic for 200 ms (or until 200 ms after I attempted to open the connection), close...
I think we do need this; it's the fastest form of storage (if backed by a fast device) and it's an important comparison point with fio and/or vdbench.
Actually, a better fix would be to change the timeout to be -1.
This should be done at some point, but is not especially high priority as the existing use of oc is well enough encapsulated.
I don't have a specific use case at the moment, so this isn't high priority. What I envision is that I want to run an existing benchmark using the benchmark-runner...
I mean the ability to run workloads without having to make any changes at all inside the benchmark-runner repository, something like pluggable modules that are dynamically loaded at runtime.
The failure in this case is external (my Elasticsearch server isn't running correctly), but the pod failure should be detected and it should error out.
Example standalone code: https://github.com/RobertKrawitz/OpenShift4-tools/blob/main/prom-extract Example metrics file: https://github.com/RobertKrawitz/OpenShift4-tools/blob/main/lib/clusterbuster/metrics-default.yaml
/sig node
Yes, this will require new documentation. I will open the website PR today.