Another question: Is there any way to (mostly) ens...
# flyte-support
g
Another question: Is there any way to (mostly) ensure or enforce a FIFO order of execution of workflows. That older workflows runs are given scheduling priority for scheduling vs newer ones?
Looking a bit online, I see that Apache Yunikorn might be a solution for this? Though in flyte docs I see that it's primarily used for training jobs as opposed to regular workflows of python tasks. Going to have a closer look at that either way
a
So when you submit a workflow for execution, propeller captures the workflow ID and adds it to the WorkQueue which is FIFO. The executor handles retries too. The order of execution may change due to multiple circumstances like failures or bottlenecks in persisting state on K8s. Some docs sections that can be helpful https://docs.flyte.org/en/latest/user_guide/concepts/component_architecture/flytepropeller_architecture.html#workqueue-workerpool https://docs.flyte.org/en/latest/deployment/configuration/performance.html#record-execution-status