These are chat archives for akkadotnet/akka.net

17th
Aug 2018
Bartosz Sypytkowski
@Horusiath
Aug 17 2018 05:18
@VladimirLuzhin so it's not a problem with ask, but with your receiver not keeping up with sending responses on time ;)
atresnjo
@atresnjo
Aug 17 2018 09:44
Anyone an idea why I am getting "Constructor on type 'Akka.Persistence.Sql.Common.Journal.DefaultTimestampProvider' not found.'" when using akka 1.3.8 with the sql provider? I've tried downgrading but still getting same error.
Roman Golenok
@shersh
Aug 17 2018 13:35
Hello, guys.
Is it good idea to use akka-net for game server backends?
Each game server is dedicated node, and cluster maybe scaled to hundreds or thousands of nodes (each node is akka-net node too).
So what problems I may met ?
Sam13
@Sam13
Aug 17 2018 13:37

@Horusiath, I fully agree with your statement that redelivery should take place when remote machine is not available for some reason.
@mwardm, I already have the actors seperated as you stated.

Due to my requirements the hardware and tasks which need to be done are given. I figured something out in meantime. I think when combining
that with improved detection of messages which have already been processed I should be fine. Thanks for your comments.

tiny hydra
@tinyhydra
Aug 17 2018 17:18
@Horusiath I was asking about reuse the other day - My actors mostly have unfortunately long-running operations, and there are a lot of them. For the past week or so, database calls have been timing out seemingly because the thread is blocked by other activities. To control this, I would like to limit my actor pool to a few instances of each to take advantage of message queueing, and by proxy, throttling this thread conflict. What would be ideal is if, when I request an actorOf, the system either returns an existing actor, or creates a new one if there's room in the pool. I've come up with my own solution, I just wanted to see if there was something like that already.
Bartosz Sypytkowski
@Horusiath
Aug 17 2018 17:26
@tinyhydra if the pool is of reasonable size, round robin router sounds like a way to go
tiny hydra
@tinyhydra
Aug 17 2018 17:28
@Horusiath I'll check that out, thanks. What would I look for to determine "reasonable size"?
Bartosz Sypytkowski
@Horusiath
Aug 17 2018 17:30
hard to say, but usually an "application works without throwing out of memory exception" is a good indicator ;)
tiny hydra
@tinyhydra
Aug 17 2018 17:44
@Horusiath memory/cpu aren't experiencing any noticeable issues.
tiny hydra
@tinyhydra
Aug 17 2018 17:51
@Horusiath I really don't know why this is happening, but I'm executing a dozen or so WMI calls per machine for a few hundred machines. When I do one at a time, everything works great, the DB writes go through instantly. When I run against even a few dozen machines, every DB write - all in their own actor - times out after 30 seconds, or longer. I've seen timeouts after 5 minutes if I set it that high.
Bartosz Sypytkowski
@Horusiath
Aug 17 2018 18:13
@tinyhydra there may be a various reasons for it - I'm not familiar with WMI characteristics. There's a risk that the client itself doesn't work well in parallel execution. If WMI client calls are synchronous you may do thread blocking.
tiny hydra
@tinyhydra
Aug 17 2018 18:20
@Horusiath I've tried running the db actors in their own dispatcher, but it doesn't seem to help. Is there a better way to isolate these threads?