These are chat archives for akkadotnet/akka.net

7th
Jun 2018
AndreSteenbergen
@AndreSteenbergen
Jun 07 2018 08:25
What settings should I adjust if I want to relax the UNREACHABLE detection? I would guess watch-failure-detector and setting threshold = 10.0 to a little hight number. I have quite high load average related to network I/O occasionally an UNREACHABLE is raised with a just a few seconds to mark REACHABLE again. At that moment PreRestart is already given, and my internal status lost (is I am correct)
watch-failure-detector {
            threshold = 20.0
            min-std-deviation = 300 ms
            acceptable-heartbeat-pause = 20 s
    }
Would that work, or too relaxed?
AndreSteenbergen
@AndreSteenbergen
Jun 07 2018 08:32
I placed it under remote, or should I place a failure detector under cluster?
Michał Dąbrowski
@defrag
Jun 07 2018 11:16
shutdown.PNG
Michal Dabrowski
@defrag2_twitter
Jun 07 2018 11:16
a question, is it normal during coordinated shutdown for last phase to last around 1 minute?
Michał Dąbrowski
@defrag
Jun 07 2018 11:17
at this moment no actor is the system is running
(lol dunno whats the mess with my gitter account :P)
Onur Gumus
@OnurGumus
Jun 07 2018 12:02
@ismaelhamed Hey Ismael, today in one occasion we have got a unique constraint violation error regarding the persistence id and sequence number. I couldn't explain how it happened. Do you have any ideas?
What is the duty of METADATA table as it is always empty for me?
Bartosz Sypytkowski
@Horusiath
Jun 07 2018 13:05
@defrag2_twitter do you have some custom shutdown stages?
Ismael Hamed
@ismaelhamed
Jun 07 2018 13:20
@OnurGumus see akkadotnet/akka.net#3468
I already fixed it in the latest release (v1.3.8): basically replaced the INSERT with a MERGE
Onur Gumus
@OnurGumus
Jun 07 2018 13:21
I think we are already using 1.3.8
@ismaelhamed it could be a testing issue too. Our tester frequently truncates the tables for testing but he is not truncating the metadata table as we always see it is empty.
Aaron Stannard
@Aaronontheweb
Jun 07 2018 13:23
that PR is still open, no?
I think I suggested that someone add some specs to put the mystery of "will this break Akka.Persistence.Sql?" to rest
Onur Gumus
@OnurGumus
Jun 07 2018 13:24
@ismaelhamed ouch ! Akka.Persistence.Oracle is at 1.3.7
Ismael Hamed
@ismaelhamed
Jun 07 2018 13:25
@Aaronontheweb We should still decide whether this is going to be solve by akka itself, or left to the plugins
Onur Gumus
@OnurGumus
Jun 07 2018 13:25
I guess I should be happy about it so this might be the issue.
Ismael Hamed
@ismaelhamed
Jun 07 2018 13:25
@OnurGumus let me know
Onur Gumus
@OnurGumus
Jun 07 2018 13:25
@ismaelhamed so I was running akka 1.3.8 with oracle 1.3.7. With combination this issue can happen right?
Ismael Hamed
@ismaelhamed
Jun 07 2018 13:26
yep
Onur Gumus
@OnurGumus
Jun 07 2018 13:26
okay good to know that I have a fix.
@ismaelhamed how about not using batchjournal . Is it safer ?
In case I have these issues keep coming, if I use non batch variant, will it help?
Ismael Hamed
@ismaelhamed
Jun 07 2018 13:31
We find the BatchingJournal times more performant than the original, and we stick with it :) Your mileage may vary, though
Onur Gumus
@OnurGumus
Jun 07 2018 13:31
@ismaelhamed no I asked in terms of this constraint violation error.
Safety is first for me, rather than performance.
Because a tiny glitch in event journaling can cause a lot of headache in my system.
Ismael Hamed
@ismaelhamed
Jun 07 2018 13:34
In that case, no. The metadata table works the same way in both implementations.
Onur Gumus
@OnurGumus
Jun 07 2018 13:34
but I always see metadata table empty.
Ismael Hamed
@ismaelhamed
Jun 07 2018 13:36
The purpose of that table is to keep track of the latest high sequence number in case you clear-up the journal
If you're not deleting, it'll be empty
Onur Gumus
@OnurGumus
Jun 07 2018 13:37
I see.
Michał Dąbrowski
@defrag
Jun 07 2018 13:53
@Horusiath no custom shutdown stages, just pinned one task at service-requests-done
and the other at service-stop. They are executing pretty quickly tho.
The long shutdown seems to be related to exit-clr=on, perhaps I dont quite
understand if i should have it on or off.
hectorcaban
@hectorcaban
Jun 07 2018 16:39
I am looking to implement akkadotnet to create load balancing, clusters solution. But I am also looking into docking/containers. What I know is that microservices can be created with dockers also. They can provide load balancing also. Should I implement both ? whats advantage if I deployed my akkadotnet solution using docker.
Bartosz Sypytkowski
@Horusiath
Jun 07 2018 18:19
@defrag not sure what could cause it, maybe it's indeed exit-clr
@hectorcaban akka.net (and basically any actor model) is more about building statefull applications, which is different paradigm than standard web-based applications
Aaron Stannard
@Aaronontheweb
Jun 07 2018 18:40
you could do load-balancing of traffic within an Akka.NET cluster though
hectorcaban
@hectorcaban
Jun 07 2018 19:15
@Aaronontheweb so to manage and scale deployments I would use docker. Everything else can be handled by Akka.NET