These are chat archives for akkadotnet/akka.net

20th
Jun 2016
Bartosz Sypytkowski
@Horusiath
Jun 20 2016 08:07
Is there any way to say to dotTrace "I'm interested in stats from this section of code, ignore the rest"?
Bart Waterschoot
@bwaterschoot
Jun 20 2016 11:02
I'm trying to test whether or not my actor receives the poison pill and terminates but can't get it to work
any ideas?
Bartosz Sypytkowski
@Horusiath
Jun 20 2016 11:32
@bwaterschoot try to use watcher.Watch(worker1.Ref); instead of watcher.Watch(worker1); (alos use worker1.Ref as target for expect terminated) as you probably want to watch an actor not the probe itself. Besides, I cannot recall if I ever used watch on probes and their refs, so I don't know if it's a valid scenario
Bart Waterschoot
@bwaterschoot
Jun 20 2016 11:45
ugh, of course that's it
thanks!
Bartosz Sypytkowski
@Horusiath
Jun 20 2016 11:48
np. I'm surprised that watch over probe itself is possible on the first place...
Aaron Stannard
@Aaronontheweb
Jun 20 2016 16:00
@alexvaluyskiy merged in your cluster spec changes
I'd never really used FluentAssertions before
I like the syntax
I think I'll start using that more often
Alex Valuyskiy
@alexvaluyskiy
Jun 20 2016 16:08
Yes, I like FluentAssertions
Muche better than Xunit.Assert, and looks more closer to scala specs
Aaron Stannard
@Aaronontheweb
Jun 20 2016 16:08
yeah
I agree
we have some of those extension methods that Hakan ported, which are part of the TestKit.Internals package
but looks like Fluent Assertions does all of that and more
Alex Valuyskiy
@alexvaluyskiy
Jun 20 2016 16:23
And about Cluster.Metrics...
Have you planned to release it in 1.1?
Aaron Stannard
@Aaronontheweb
Jun 20 2016 16:24
nah, decided on a middleground approach
remove the bits from Akka.Cluster
so those APIs aren't included
Alex Valuyskiy
@alexvaluyskiy
Jun 20 2016 16:24
Cluster.Metric is not only - moving code from Cluster package to a separate
Aaron Stannard
@Aaronontheweb
Jun 20 2016 16:24
and stick them into a new library
Alex Valuyskiy
@alexvaluyskiy
Jun 20 2016 16:24
We should create Routers for that
Aaron Stannard
@Aaronontheweb
Jun 20 2016 16:24
which probably won't be released
in 1.1
if we have time to fit it in, great
if not
it's something we could cut in order to ship them
err, ship 1.1 on time
(well, not on time)
damnit I need more coffee
voltcode
@voltcode
Jun 20 2016 17:03
@Horusiath I used profiling API in the past with success, although not with dotTrace but with VS profiler
Alex Valuyskiy
@alexvaluyskiy
Jun 20 2016 17:23
@Aaronontheweb I prefer to make Cluster Tools/Sharding stabilization is more preferable than Cluster Metrics
Cluster Singleton is already stable akkadotnet/akka.net#2102
Bartosz Sypytkowski
@Horusiath
Jun 20 2016 17:45
@Aaronontheweb some help here?: [ERROR][2016-06-20 17:45:00][Thread 0004][akka://cluster/system/endpointManager] Method 'Configure' in type 'Akka.Remote.Transport.Helios.TcpServerHandler' from assembly 'Akka.Remote, Version=1.0.8.24, Culture=neutral, PublicKeyToken=null' does not have an implementation.
Aaron Stannard
@Aaronontheweb
Jun 20 2016 17:46
upgraded to Helios 2 by accident
downgrade to 1.4.1
what do we need to do with the Nuspec to lock the version\
prevent nuget from upgrading
voltcode
@voltcode
Jun 20 2016 17:48
@Horusiath I noticed I missed the link before, sorry https://msdn.microsoft.com/en-us/library/bb514149.aspx
Bartosz Sypytkowski
@Horusiath
Jun 20 2016 17:49
np. I've found what I needed anyway ;)
Aaron Stannard
@Aaronontheweb
Jun 20 2016 18:02
well, I can do a spot-release onto master that fixes the versioning specification inside FAKE
or we can just wait for 1.1
I'd opt for doing a quickfix for nuget
locks the current released packages to 1.4.1
prevents the upgrade
I'll need to dick around with FAKE to make that work but I'm sure it's doable
Aaron Stannard
@Aaronontheweb
Jun 20 2016 18:18
@alexvaluyskiy think I've found the source of some of the issues with remote deathwatch
issue is that inside Akka.Remote that are some areas where we're not handling the UID consistently
Aaron Stannard
@Aaronontheweb
Jun 20 2016 20:15
@rogeralsing ouch, just found a subtle bug in the UID system
since we're using incrementing counters for actor UIDs
what happens if we shutdown a remote actor system, restart it, and recreate the actors in the same order every time?
the UIDs are the same
so we should be using a random number, just like on the JVM
Aaron Stannard
@Aaronontheweb
Jun 20 2016 20:24
watched an instance of this happen in a multi-node spec where we simulated restarting a node
that's not the reason why the spec was failing
reason was that we're not including the UID in a deathwatch operation somewhere inside Akka.Remote, but I saw the actor get re-created in a new actor system with the same UID
either way, changed it to random now