@Aaronontheweb hmmm... I've seen the multi-node test runner hang twice on this:Step 1/1: Writing test state to: D:\BuildAgent\work\49b164d63843fb4\TestResults\Akka.Cluster.Tests.MultiNode-635834644789423697.json including right now literally just writing to a file
logged #1433 for it
just saw it happen again
@chillitom in order for serialization to be fully transparent, it must be fully qualified
all statically typed, in other words
although I thought the Google Protocol Buffer envelope around the JSON blob had the type name burned into it
so it may not be necessary to do it again inside the JSON
that's a better question to ask @rogeralsing
@chillitom we need to support subtype serialization. e.g. if you have a property of type animal but the value is of type cat, then the deserializaed value needs to be materialized back as cat
@rogeralsing thanks :+1:
also if you have arrays of objects, like we do for constructor arguments for remote deployment, then we need the type for all the args
okay, makes sense
do you think it could be exposed as an option? we are simply using f# discriminated unions as 'Event types, would be nice if they could be kept short and sweet
in the postgres adaptor theres a comment about restricting the ticks column
@chillitom in general, when you use persisted events, never let other party decide how should they be serialized
use your own serializer instead
cc @davidflerlage regarding Postgres plugin
was wondering if someone could explain "restricting created_at_ticks to the range [0,10) to ensure that there are no precision differences in the opposite direction"
I haven't applied such a constraint in the MySQL adaptor as unsure of the reasons behind it
@Horusiath thanks for the advice, I'll do that
man, I can't get a single build to pass right now with all of these racy spec failures
even a PR for fixing one of them! :p
only happens on the el cheapo Azure boxes
and not my dev machine
Has anyone seen an issue in Akka.Persistence where the state update falls behind writing to the persistence store?
it seems UpdateState is falling behind on hight throughput
Persist claims this shouldn't happen (obv)
Hi guys, I need some help with IIS and actor system. Last night, I receive Information regarding "Remoting shutdowon" from my asp.net application, and then my windows service which connects through akka remote throws AssociationError. The problem is that once the remote shutdown in asp.net application, it doesn't come back until I hit the IIS with a http request. Is there anyway I can make sure that the actor system will be restart in this case?
yep, so this is a common issue with IIS
you basically have to configure IIS to never, ever, under any circumstances, recycle your ASP.NET application if it goes idle
@Aaronontheweb I try to set the idle time to zero, but it still happens...
state changes can't happen our of order
@Aaronontheweb Maybe I should write another windows service which request the http end pint X time a day?
might be another IIS setting overriding it
which would be weird
protip though: your Windows Service should only push to the web app if people are actually using it
if no one is using it and you do a server-push, nothing happens
so I'd design your service relationship that knowing that
that the two services don't always have to both be up - IIS does weird shit and is yet another example of Microsoft trying way too hard to think for the developer
so rather than hack IIS, design your Windows Service so it's still "up" and not constantly trying to call someone who isn't home
that's just my 0.02
@Aaronontheweb re #1433 is anything written at all? or is the hang on first write?