2020-10-31 01:14:04 UTC - Brendan Doyle: I'm looking into making some minor optimizations to the load balancer algorithm. For home invokers, is there a reason that every action that shares the same home invoker has the same step size performance wise? I understand the reasoning for having the different step sizes, but in my head I'm thinking that the step size should be decided at the action level and not per invoker. Can be done by re-hashing somehow after the initial hash. But it seems intuitive to me so that spillover of two high load functions competing with each other go to two different invokers (or at least have a chance to)? https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1604106844183200?thread_ts=1604106844.183200&cid=C3TPCAQG1 ---- 2020-10-31 08:12:56 UTC - tycho: Hi all, I managed to deploy and run with kafka + couchdb and it works fine, I managed to expose the playground on a public address (as I deployed the local install on a server so my colleagues also can have a go) and that works well as well. I have written different functions, connected to our psql db and more, written triggers etc. However, I cannot figure out one thing (I guess there will be more) and the only online info about it seems to be polling; if I call a function ?blocking=false I receive an activationId as I should do. Now I can poll this activationId and that works but is quite... (even though the only tutorials online, like the adobe one and some guys' blogpost do it that way!) not so good. We (as in openwhisk users) have Kafka running which (if i read the architecture page correctly) seems to get the result data; so, the question is; how do I subscribe to the result data instead of polling it? I was expecting something easy for that like ?blocking=false&webhook=https://.... I guess it can be done by creating a trigger, firing it from the longrunning function, having an action called and have that action call a webhook. That seems ... also not very nice for something that we would need *all* the time (firing up an instance for something Kafka was actually created for...) so I hope i'm missing something trivial here! Thanks! And keep up the great work! https://openwhisk-team.slack.com/archives/C3TPCAQG1/p1604131976191300 ----
