@Aaronontheweb good article, have you actually used akka.net cluster in prod with good success? ... would love to that, but those intermittent tcp errors are not giving me a warm and fuzzy feeling
i have looked at the comments and issues today, looks like others are running to some of those as well
This message was deleted
@tstojecki senor @annymsMthd has. You have to bear in mind that a lot of the issues that come up with Akka.Cluster when you're initially launching it have to do with two nodes racing to connect with each other, dropping the connection attempt when they realize that two are occurring at once (which is illegal,) and then waiting at random intervals to attempt to connect again.
Cassandra, Riak, and other clustered DB systems run into the same issue
and it's something that comes up more often in dev and test than production
that being said, eliminating intermittent TCP connection errors is something that we're still working on
1.0.2 resolved a lot of that
but there's still a lot more work to do
I'm just about done with the R&D I had to do to develop Petabridge's training courses (gotta pay the bills,) shifting back to full-time R&D on Akka.Cluster and Helios next week
I've personally used Akka.Remote in production under what I would consider to be very heavy workloads
great to hear you guys are aware and on top of that
yeah, I'm anal about the quality control on Akka.Cluster - but a lot of the issues are really subtle
i.e. we realized that running the heartbeat system on top of the TPL had devastating consequences under even light loads
moved that to the dedicated scheduler and resolved a lot of heartbeat issues
and I'm pretty sure there are still some issues inside the EndpointManager
lol. just accidentally went to akka.net instead of getakka.net
I sent a mail to the domain owners last week
No reply though
in case they want money for domain ownership change we can organize crowdfunding campaign :)
Hello. A quick check. Did somebody has experience with Azure webjobs and Akka. I want to run the my Akka System inside a WebJob that will be triggered by a queue item. I am not sure if Azure webjobs creates separate processes per job. dequeue Item. I don't want to have two Actor System running in parallel.
If that is the case, you can use akka.remote to communicate with a shared system.
Webjob1 -> shared remote system <- webjob2
That is good to know. I found the answer. There is a single instance of WebJob per website Instance. You can even mark the webjob as singleton in case you have multiple instances.
I feel that i need to do some learning/reading/experimenting before going in akka.remote
How do you guys get your IActorRefs from "outside" the "actor system". "Global" reference to the actor system?
what do you mean?
how to compose actor system with the rest of the application?
I don't have a particular example, but lets say you need an actor from a MVC controller or some other part that is not directly part of the actual actor system
I'm just getting started and have a basic understanding of actor selection etc. but just want to know how others deal with, yeah composing the non actor part of the application to work with the actor system :)
one of the most popular ways is to create static class container for actor system, initialize it in your Global.asax or OWIN Startup class and close when Application ends
Yeah I was leaning that way, thank you for confirming. I'll try that approach. Do you send IActorRefs around too or do you always go through the "root" using actor selection?
Thanks, I think I already read that one, but I'll have another look :) thank you for your time.
Quick poll of gitter chat: have any of you heard of "extend-only design" before?
I'm writing a section on message-versioning strategies for our clustering course and am having trouble finding any actual literature on it - it's a schema / message versioning technique I've used for like 5 years that really makes upgrades trivial to do
and rollbacks too, I might add
I just assumed that this was a commonplace thing
but judging by what I can find on Google, that assumption might be off