flytestdlib
flytestdlib copied to clipboard
[wip] FairQ: An implementation of K8s Workqueue
- The implementation aims to provide equal worker allocation across all namespaces. This avoids one namespace from taking over all workers and thus increasing the latency to process a namespace with lower load.
- The queue is thus fair for every namespace. IF somenamespace has skewed work duration (larger) then it will not inversely impact the reconcilation latency for objects in another namespace
TODO: workqueue metrics support