[slack-digest] [2022-05-09] #general

[OpenWhisk Team Slack <ra...@apache.org>]

2022-05-09 21:24:59 UTC - Brendan Doyle: On the new scheduler, the ability to track a namespace's container and action usage in metrics is awesome. Is there any way to track total user memory used by a namespace and action yet since a namespace can have actions of varying memory provisioned. You can obviously do this yourself by adding up all of the containers for each action and the memory setting, but it would be really nice if this was being tracked. I'll create an issue to track it if not

Other than that the new scheduler seems awesome to me so far. I am wondering how we can lower the aggressiveness of spawning containers the first time the scheduler sees an action. Some operators may not care about imposing a bit of latency initially in exchange for the scheduler learning the traffic pattern and not overwhelming the fleet / the namespace's available concurrency too quickly. What would be the best existing config to adjust this right now? I know there's the activation staleness in the queue but that's hardcoded at the moment at 100ms
https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1652131499780739?thread_ts=1652131499.780739&cid=C3TPCAQG1
----