These are chat archives for akkadotnet/akka.net

25th
Apr 2018
Bartosz Sypytkowski
@Horusiath
Apr 25 2018 04:12
@davzucky I've got some initial boilerplate code for akka.http, but it's months of work from being done.
Hyungho Ko
@hhko
Apr 25 2018 09:16
@Horusiath thank you
I have some problems about ActorSelection.
my deamon service have memory leak.
this is test solution for memory test.
and this is summary for it.
ActorSelection_MemoryTest_Summary.png
our test solution scenario is very simple.
Test client program only call ActorSelection and exits.
Hyungho Ko
@hhko
Apr 25 2018 09:21
on a per-process basis.
after build, execute Server process and ClientExecutor process in Client bin folder
Boban
@bobanco
Apr 25 2018 09:34
@hhko i dont think its memory leak, by default dotnetty is using pooled buffers, @Aaronontheweb might confirm whats the default in akka
Hyungho Ko
@hhko
Apr 25 2018 09:36
DotNetty version memory continues to grow up to 1G
as as result
DotNetty version ActorSelection always fails.
How to control pooled buffers size?
AndreSteenbergen
@AndreSteenbergen
Apr 25 2018 09:44
Hi guys, is it possible to get a list of all actors in the system? I am doing some crawling, but I want to get rid of all actors once a domain is fully crawled.
AndreSteenbergen
@AndreSteenbergen
Apr 25 2018 09:51
Is there any way I can check the active actors? I think I have send stops and send poison pills to the correct actors. But I would like to know for sure ...
PetabridgeCmd support "actor hierarchy"
on run-time.
AndreSteenbergen
@AndreSteenbergen
Apr 25 2018 10:02
Thanks
Hyungho Ko
@hhko
Apr 25 2018 10:33
Akka.Remote.Transport.Helios 1.3.6 also has same situation about Continually increasing memory up to about 2G.
at that time, ActorSelection always fails.
Bert Lamb
@BertLamb
Apr 25 2018 13:21
I am finding myself having writers block when naming Actors, is there some sort of helpful naming guide out there? My brain just wants to start calling things Managers or Coordinators and that just feels dumb.
Stijn Herreman
@stijnherreman
Apr 25 2018 13:29
@BertLamb my top-level actors are named ...Coordinator
nothing wrong with that, if they're actually coordinating things :)
some are also named ...Supervisor
lower level actors are named after what they represent, e.g. the Specification actor accepts a GetSpecification message, the DomainTest actor accepts a PerformDomainTest message, etc
I found it very useful to draw a schema of the actor hierarchy, it helped with naming things
Bert Lamb
@BertLamb
Apr 25 2018 13:34
:thumbsup:
David Zucker
@davzucky
Apr 25 2018 14:33
@Horusiath Let me know if you want any help. I'm looking to get that for a project and I would be happy to help you.
Vagif Abilov
@object
Apr 25 2018 15:51
Hi guys, we reconfigured some routers to be cluster-aware and now we are no longer able to send messages to their routees by using their addresses. For example, when a routee was non-clustered, we could retrieve its path (something like user/xyz/$b), and it was possible to send a message to the specified path. When the router is cluster-aware, routees have different naming schema (something like user/xyz/c1), and when a message is sent directly to such address, it doesn't get delivered.
I wonder what is a right way to send a message to a specific child of a cluster-aware router. Or should such messages with direct address be avoided since routees can be reconfigured and moved between nodes, so a reference to a specific host may become invalid?
If so, what is the right way to implement callback to children of cluster aware routers, so they can receive messages that will bypass router and will come directly to them?
Michal Dabrowski
@defrag2_twitter
Apr 25 2018 16:28
Hi guys, anyone knows perhaps if its by any chance doable to connect to azure web job running akka
from web app? Connections between multiple webjobs works fine, tho from web app it just fails silently.
Arjen Smits
@Danthar
Apr 25 2018 16:52
sometimes.. github's git implementation, really makes me angry
Arjen Smits
@Danthar
Apr 25 2018 17:10
67 rebase commits to fix, just because i want to change the base of a branch with 1 commit
facepalm
Ok i dont know why. But git wont allow me to rebase a dev based branch onto v1.4 without integration all the changes between dev and the 1.4. Its like totally ignoring the common commit of my branch
im nuking it, and manually reapplying the fix in a new branch. Takes alot less time then manhandling git
Aaron Stannard
@Aaronontheweb
Apr 25 2018 17:16
@object ah, you're talking about cluster pool routers?
or is this a cluster group router?
lol @Danthar
I know that feel
Vagif Abilov
@object
Apr 25 2018 18:27
@Aaronontheweb cluster pool routers.
Aaron Stannard
@Aaronontheweb
Apr 25 2018 18:27
ok got it
that makes sense then
I wonder what is a right way to send a message to a specific child of a cluster-aware router. Or should such messages with direct address be avoided since routees can be reconfigured and moved between nodes, so a reference to a specific host may become invalid?
so you're affected by this bug...
(grabbing the issue)
Vagif Abilov
@object
Apr 25 2018 18:29
Aha, @pmbanka who discovered the problem suspected that it might be a bug.
Aaron Stannard
@Aaronontheweb
Apr 25 2018 18:30
akkadotnet/akka.net#1544
it's an ooooooold issue
but we've never really attempted to fix it
our ActorSelection code doesn't go a good job dealing with the virtual path container we use for remotely deployed actors
so we'd need to modify the code we use for resolving ActorSelections inside Akka.Remote
we really should fix it, but it just hasn't been all that urgent
I take it this is a big pain in the ass for you guys?
if so, I can get on it - may not be for another week or so, but I'll put it on my to-do list
Vagif Abilov
@object
Apr 25 2018 18:33
Both Pawel and me sent some PRs in the past but I have no idea whether it's feasible for us to figure out how to fix it ourselves.
Aaron Stannard
@Aaronontheweb
Apr 25 2018 18:34
Akka.Remote is a bit of a bitch in general to work on
since it has to do all of the funky location transparency / address translation stuff
I'll take it on and see if we can get it into a nightly soon
Vagif Abilov
@object
Apr 25 2018 18:35
And it's a kind of pain in the ass since it prevents us from using some routers in a cluster.
Aaron Stannard
@Aaronontheweb
Apr 25 2018 18:35
alrighty, I'm on it
made it Github-official
Vagif Abilov
@object
Apr 25 2018 18:36
🙏
Aaron Stannard
@Aaronontheweb
Apr 25 2018 18:36
subscribe to that issue I linked to for updates
Vagif Abilov
@object
Apr 25 2018 18:36
Thanks Aaron!
Aaron Stannard
@Aaronontheweb
Apr 25 2018 18:36
I'll add some failing unit tests and work backwards
happy to help
always love a good puzzle
plus for some reason
closing a really old issue always makes me feel good
@defrag2_twitter > Hi guys, anyone knows perhaps if its by any chance doable to connect to azure web job running akka
from web app?
IIRC
Azure Web Apps lock down the ports you can listen on really aggressively
so our normal DotNetty transport can't bind to anything
we've talked about building a SignalR transport as a work-around for Azure Websites and Azure Apps
which would honestly be a pretty fun project to work on, speaking for purely self-interested reasons :p
other possibility would be an HTTP/2 streaming transport
either one of those should work in an environment where you can't rely on sockets
Vagif Abilov
@object
Apr 25 2018 18:40
Reading the history, it's soo old!
Vagif Abilov
@object
Apr 25 2018 18:49
Kind of strange that not many developers hit this bug. Makes me wonder if there are more preferred ways to achieve the same result. Pawel suggested use of distributed PubSub but I am not sure whether this would be the right pattern if we want to send a message to a specific remote actor which happens to be a routee of a clustered pool.
Aaron Stannard
@Aaronontheweb
Apr 25 2018 19:10
in general remote deployments are relatively rare in Akka.Cluster usage in the wild
even rarer is needing to communicate with those remotely deployed actors via ActorSelection
which is why I think it hasn't been raised higher up the flag pole until now :p
but we've been aware of it and it does look like @marcpiechura put some time in a couple of years ago on it