set the channel description: Find in this channel conversations that result from the Flyte School sessions.

Hi everyone ICYMI Flyte School ep. #4: Q&A 1. can we use any FSSPEC compliant Filesystem ? a. While Flyte 1.5 introduced experimental suport for the fsspec API, currently only S3-compliant options can be used as blob storage to store Task inputs/outputs and other metadata. 2. How long the cached results is retained on blob storage? a. By default cache never expires, but this behavior can be controlled with

max-cache-age

. “max-cache-age” defines when Flyte will stop using the values but, if there are cached outputs in object storage that need to be lifecycled, that needs be taken care of by bucket lifecycling policies - you don’t get that OOB with Flyte. One way to cleanup might be to set

max-cache-age

to 7d, and lifecycle objects in your bucket after 14 days. cc @Hema Jayachandran 3. We can trigger Runs by CRONS, but can we Run a Workflow when another one finishes ? a. You can achieve this by chaninig workflows using the

>>

operator. Docs:https://docs.flyte.org/projects/cookbook/en/latest/auto_examples/advanced_composition/chain_entities.html 4. is it sensible to poll an external service in a workflow? i read it's currently an anti-pattern but not sure why a. Technically, nothing prevents you from doing it. but definitely disable caching when doing it since the results may not be deterministic. Also, you could explore using sensors. cc @Alexander Walther -- Recording: Next session: Sep 27, 9:00 a.m. PT. We'll cover Flyte deployment options and will perform a live install Thank you all!

Hi David, Thank you for getting back on this gratitude thank you. So we have a retention policy on s3 which is 28 days but if I understood you right, then setting max-cache-age to 28d would help.

One way to cleanup might be to set

max-cache-age

to 7d, and lifecycle objects in your bucket after 14 days.

I quite didn’t understood though why lifecycle has to be greater than max-cache age. Or it was just an example ?