Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Apr 06 22:45

    Aaronontheweb on dev

    sharding - release shard lease … (compare)

  • Apr 06 22:45
    Aaronontheweb closed #4369
  • Apr 06 22:45
    Aaronontheweb milestoned #4369
  • Apr 06 22:45
    Aaronontheweb labeled #4369
  • Apr 06 22:45
    Aaronontheweb commented #4363
  • Apr 06 22:31
    yesh-nadella commented #4363
  • Apr 06 22:05
    Arkatufus commented #4368
  • Apr 06 19:55
    zbynek001 opened #4369
  • Apr 06 17:38
    Aaronontheweb assigned #4368
  • Apr 06 17:38
    Aaronontheweb labeled #4368
  • Apr 06 17:37
    Aaronontheweb milestoned #4368
  • Apr 06 17:37
    Aaronontheweb labeled #4368
  • Apr 06 15:07
    PiotrJustyna opened #4368
  • Apr 03 22:24
    huzaifak opened #4367
  • Apr 03 21:45
    Aaronontheweb commented #4344
  • Apr 03 05:55
    Lutando commented #4344
  • Apr 02 17:23
    huzaifak commented #4303
  • Apr 02 17:23
    huzaifak closed #4303
  • Apr 02 02:13
    ShawnYun commented #4366
  • Apr 01 17:18
    huzaifak commented #4303
Damian Reeves
@DamianReeves
Cool.. Set the seed node and now I'm in business. Thanks
Bartosz Sypytkowski
@Horusiath
np :)
Damian Reeves
@DamianReeves
Is there a simple way to programmatically add roles to a Cluster node
?
Peter Bergman
@peter-bannerflow
I think this question has been posted before but I cannot find the answer so here it goes, if I do an ActorSelection, can I then somehow check if that leads to an actor or not? If that makes sense...
Arjen Smits
@Danthar
@peter-bannerflow you can use the ResolveOne method on an actorselection
Peter Bergman
@peter-bannerflow
Ah yes, or actually I only want to do this in a case where the ActorSelection leads to a child of Self, so I guess I can just do Context.Child("name").IsNobody()
Arjen Smits
@Danthar
or send an Identify message, which will return an ActorIdentity response.
ah yes, if you want to know if a child exists checking for the nobody status is faster
Peter Bergman
@peter-bannerflow
Cool, thx
Sean Farrow
@SeanFarrow

@damianReeves are you talking about adding new nodes at the same time as roles, or it is just roles to a node?

+

All Conversations

Activate or click this element to toggle the profile menu

Search

Favourites

People

akkadotnet

Home

Billing

Get Gitter Apps

Sign Out

Add a room

akka.net

101

Leave

Hide

AkkaStreams

Leave

Hide

All Rooms

dotnet-core

akkadotnet.github.com

Wire

Peter Bergman @ akkadotnet/akka.net

Cool, thx

Arjen Smits @ akkadotnet/akka.net

ah yes, if you want to know if a child exists checking for the nobody status is faster

Damian Reeves
@DamianReeves
roles to a node
i.e. I can call cluster.Join to join a cluster but how do I say what roles I provide programmatically?
verilocation
@verilocation

Hi... I hope someone can help. Im currently evaluating Akka.net for suitability for an internal rewrite. Its going well so far but I am currently looking at edge cases. I have setup a Cluster with remote deployment but want to be able to resend a message on failure. Is there anyway I can detect this and retry?

I have tried to look to see if I can read messages off the DeadLetter mailbox and retry from there but didn't have much luck trying to find out how to read from that mailbox either.

Damian Reeves
@DamianReeves
You could take a look at AtLeastOnceDeliveryActor http://getakka.net/docs/persistence/at-least-once-delivery
That allows you to require message acknowledgement
verilocation
@verilocation
Thank you very much... Persistence isnt fully release though yet is it?
Damian Reeves
@DamianReeves
no that is planned to come out of beta in 1.5
verilocation
@verilocation
ok great thanks, thats this circa year isnt it?
(end of)
Damian Reeves
@DamianReeves
Thats the plan
verilocation
@verilocation
:thumbs_up:
:thumbsup:
thanks for your help / quick reply
Damian Reeves
@DamianReeves
np
verilocation
@verilocation
Me again.. Sorry.. Quick related Q... Is there anyway to tell if a Tell or Ask failed, so that I can retry myself? I'd be happy if I could, a little like RabbitMQ, retry 3 times before giving up. Anecdotally, it seems that when you kill a node it takes 3 dropped messages for the rest of the cluster to figure out it has died and to then stop sending it messages
Andrey Leskov
@andreyleskov
You can try Ask with timeout or expect a Failure in response
verilocation
@verilocation
If I don't specify a Timeout will it wait indefinitely? Can I specify an app level timeout
Andrey Leskov
@andreyleskov
Yes, we will be waiting forever. I'm not aware about specifying default timeout possibility 8 (
Bartosz Sypytkowski
@Horusiath
@DamianReeves you roles are fixed at cluster startup, you cannot change them when the system is running
Arjen Smits
@Danthar
@verilocation you can specify the global ask timeout in your hocon config.
Aaron Stannard
@Aaronontheweb
@mmisztal1980 you'd want to use a Docker network
and using something like Docker Swarm or Mesos or etc, pull the addresses of other registered containers on that network acting like seeds
I've been using good old-fashioned Docker links to do that too
which isn't great for production stuff IMHO
fine for testing
Docker network is a NAT that can span multiple Docker Machines
so that's how you get containers on different boxes talking to each other
whereas a link, I think can span the network but it's a point-to-pint descriptor you have to pass in when launching the image
@Danthar lol yes... there are a bunch that need that
the design patterns one too
@raskolnikoov > is it possible to schedule the actorsystem scheduler to call an actor at random times every day?
hmmm....
not by default
but what you could do: send a message to an actor at fixed intervals and have the actor randomly decide whether to handle it or not
achieves the same thing, but backwards - you could put the "randomizer" in as a kind of facade actor
and have the randomizer, when it chooses to accept the message, pass the message along to the real actor downstream
Aaron Stannard
@Aaronontheweb
other possibility is instead of using "tell repeatedly"
you could just generate a random schedule of single "tell once" calls to the actor
and do that once a day or something