These are chat archives for akkadotnet/akka.net

14th
Feb 2018
Garrard Kitchen
@garrardkitchen
Feb 14 2018 15:10
I would love to hear from anyone who has managed to get clustering working across to docker hosts (not containers on a single host) on AWS ECS. I thought it would be as simple as keeping a list of seeds in dynamodb using lambda and when node (container) comes online it builds seeds list from calling lambda. Connectivity of containers within a single host works fine but issue is when you cross host boundary (required for HA). I'm using bind-hostname to hold host ip address but not sure if this hocon setting is implemented. I have seen mention of bind-port but no way of ascertaining dynamic host port for container from within container itself. I have thought about writing small agent to run on windows host that periodically talks to aws api to obtain ports for all containers and push this into file on volume that is being shared between host and container and have container access this file to get dynamic port on node start up but prefer to hear from somebody that has experienced the same pain first.
AltcoinsBattle
@AltcoinsBattle_twitter
Feb 14 2018 15:55
Hey guys, does akka have a way to monitor what is going on inside? I have almost empty system that gradually increases the cpu usage.. 5 min and it jumps to 25% CPU .. around 10 min to 50% .. and stay on that level
Joshua Garnett
@joshgarnett
Feb 14 2018 15:57
@garrardkitchen to get it working on ECS, I ended up creating an overlay network on my hosts with flanneld, so each container had a unique ip and could always be addressed using the same port
Paweł Bańka
@pmbanka
Feb 14 2018 16:07
Hi! I just noticed that Akka.TestKit.NUnit current version is 1.3.2 (https://www.nuget.org/packages/Akka.TestKit.NUnit/). Is this on purpose, or an omission during releasing 1.3.4 packages?
Razvan Goga
@razvangoga
Feb 14 2018 16:48
Hi! I have a bunch of top level actors (created directly from the system) just to encapsulate some one-off background activities (so my hierachy is 1 level deep from /user but very broad). How should the supervision work in this case ? Can I configure the /user actor's supervisions strategy ? Should I have an intermediate actor just for the supervision? Should I just hand roll my error handling in the actors? Which is the recommended practice for this scenario? Thanks, Razvan
Onur Gumus
@OnurGumus
Feb 14 2018 17:03
@razvangoga Yes you should have your global parent actor
It also serves as a global stopping point for your application
Razvan Goga
@razvangoga
Feb 14 2018 17:55
@OnurGumus got it, thank you.
Garrard Kitchen
@garrardkitchen
Feb 14 2018 19:55
Thanks @joshgarnett, this sounds complicated. And it works ok on windows hosts yes?
Joshua Garnett
@joshgarnett
Feb 14 2018 21:21
I haven’t tried on Windows hosts, sorry
Onur Gumus
@OnurGumus
Feb 14 2018 22:39
How do you test a PersistentFSM ?
There is a TestFSMRef but this cannot be used for PersistentFSM