Stratara.Testing.Orleans
4.4.1
Prefix Reserved
See the version list below for details.
dotnet add package Stratara.Testing.Orleans --version 4.4.1
NuGet\Install-Package Stratara.Testing.Orleans -Version 4.4.1
<PackageReference Include="Stratara.Testing.Orleans" Version="4.4.1" />
<PackageVersion Include="Stratara.Testing.Orleans" Version="4.4.1" />
<PackageReference Include="Stratara.Testing.Orleans" />
paket add Stratara.Testing.Orleans --version 4.4.1
#r "nuget: Stratara.Testing.Orleans, 4.4.1"
#:package Stratara.Testing.Orleans@4.4.1
#addin nuget:?package=Stratara.Testing.Orleans&version=4.4.1
#tool nuget:?package=Stratara.Testing.Orleans&version=4.4.1
Stratara.Testing.Orleans
Derived. The behaviour described here is specified under
openspec/specs/. Those specifications are the source; this page explains and illustrates them.
Run the Orleans execution model in a test's own process. ExecutionModelTestHost starts one
silo clustered with itself, with reminders and the grain directory in memory, and the framework's
real write stack, portable commit-order reader, checkpoint store and intent store on an in-memory
SQLite database. Every period the model keeps as a reminder or a poll is shortened to seconds, so a
projection applies and a timer fires within a test. No cluster, no broker, no database server, no
Docker.
Builds on Stratara.Testing.EntityFrameworkCore.
Why run the real grains?
Because the registration is what a test should prove. Your handlers, projections, sagas, processes
and timer ports are registered with the calls production uses — AddStrataraAggregateGrains,
AddStrataraProjectionGrains, AddStrataraSagaGrains, AddStrataraDurableTimers,
AddStrataraOrleansCommandDispatcher — and run in the same grains, through the same mediator
pipeline, reading the store in commit order from a checkpoint.
Example
await using var host = await ExecutionModelTestHost.CreateAsync(services => services
.AddAggregatesFromAssemblyContaining<Account>()
.AddCommandHandlersFromAssemblyContaining<OpenAccountHandler>()
.AddProjectionsFromAssemblyContaining<BalanceProjection>()
.AddStrataraAggregateGrains()
.AddStrataraProjectionGrains());
await host.DispatchAsync(new OpenAccount(accountId, 100m)); // runs in the account's activation
await host.WaitForReadersAsync(); // every reader is at the store's head
Assert.Equal(100m, balances[accountId]);
Contents
ExecutionModelTestHost—CreateAsync(configure, options),DispatchAsync,WaitForReadersAsync,Timers,SeedAtHeadAsync,ResetAsync, the presetSessionandServices.IAsyncDisposable.ExecutionModelTestHostOptions— the periods (PollInterval250 ms,ReminderPeriod1 s,IntentGrace2 s,DrainPollingInterval1 s), the partition count (4), the start timeout, andBeforeStart, which runs before the silo starts — where a test appends history and seeds the readers at its head, as a deployment does while no silo runs. A value a test sets through a registration's own settings wins over the host's.StrataraTestReadDbContext— the read context that holds the checkpoints.
Share a host across a test class
Creating a host takes a few seconds. Share one through a class fixture and reset it between tests where one must not see another's state:
public sealed class ExecutionModelFixture : IAsyncLifetime
{
public ExecutionModelTestHost Host { get; private set; } = null!;
public async ValueTask InitializeAsync() =>
Host = await ExecutionModelTestHost.CreateAsync(services => services /* the roles */);
public ValueTask DisposeAsync() => Host.DisposeAsync();
}
ResetAsync runs the execution model's reset port on this host, where the silo keeps running: the timers
and the grain directory's entries go, and every registered reader is put at the store's head. The next
test's facts are applied, the last test's are not applied a second time, and WaitForReadersAsync returns
at once. The read models themselves are not emptied — empty or rebuild them in the fixture where a test
needs them clean.
Where the host stops
- One silo. Membership, reminders and the directory are in memory and not shared. What happens across silos — a kill, a role split, a takeover — is an integration test against real infrastructure.
- Test projects only. Referencing the package from a project that is not a test project fails
the build with
STRATARA1001;StrataraAllowTestSupportOutsideTests=trueopts out, as the sample does.CreateAsyncrefuses to run whereDOTNET_ENVIRONMENTorASPNETCORE_ENVIRONMENTstates an environment other thanDevelopment.
Dependencies
Stratara.Orleans,Stratara.Orleans.EntityFrameworkCore,Stratara.Testing.EntityFrameworkCoreMicrosoft.Orleans.Server,Microsoft.Extensions.Hosting
| Product | Versions Compatible and additional computed target framework versions. |
|---|---|
| .NET | net10.0 is compatible. net10.0-android was computed. net10.0-browser was computed. net10.0-ios was computed. net10.0-maccatalyst was computed. net10.0-macos was computed. net10.0-tvos was computed. net10.0-windows was computed. |
-
net10.0
- Microsoft.Extensions.Hosting (>= 10.0.11)
- Microsoft.Orleans.Server (>= 10.3.1 && < 11.0.0)
- Stratara.Orleans (>= 4.4.1)
- Stratara.Orleans.EntityFrameworkCore (>= 4.4.1)
- Stratara.Testing.EntityFrameworkCore (>= 4.4.1)
NuGet packages
This package is not used by any NuGet packages.
GitHub repositories
This package is not used by any popular GitHub repositories.
A release about errors that were not errors. Where two writers insert the same row at once — two
deliveries forgetting a deleted tenant, two activations writing a partition's first checkpoint, two
hosts importing one API key — the framework already recovered, but EF Core had logged the losing
statement at Error first; on PostgreSQL and SQLite the insert now skips a row that is already there.
And a RabbitMQ bus whose broker went away first no longer reports a warning for every subscription it
stops, nor a failed handler for a message it merely could not acknowledge. No migration and no
configuration change is needed.
### Fixed
- **Forgetting a deleted tenant leaves no error in the log when two deliveries race.** A tenant's two
deletion facts arrive in two bundles, so two deliveries of one projection can record the same tenant
at once. The losing insert was caught and the tenant recorded all the same, but EF Core had already
logged the failed statement at Error (`20102` and `10000`), so a customer deletion could leave two
errors in a projection worker's log. On PostgreSQL and SQLite the tenants are now recorded with an
insert that skips a row already there; on other providers the store still retries, and EF Core still
logs the statement that lost.
- **A projection checkpoint's first write, and an API key two hosts import at once, leave no error in
the log.** Two activations of one reader that overlap during a failover can both write a partition's
first checkpoint, and hosts that seed the same key as they boot import it at once. The writer whose
insert lost recovered — it took the other's checkpoint over, or adopted the other's key — but EF Core
had already logged the failed statement at Error. On PostgreSQL and SQLite the insert now skips a row
that is already there, as the record of forgotten tenants does; a checkpoint that another writer
inserted and removed again before it could be taken over is refused with an
`InvalidOperationException`. An import takes that path only where it is the only write on its
context, and then writes the key and its membership in one transaction of its own, outside
`SaveChanges` and its interceptors. An import on a context with unsaved changes or inside a
transaction of the caller's saves through `SaveChanges` as before, and so does a table a consumer
widened with columns the row does not carry, and every other provider: there the writer still
recovers, and EF Core still logs the statement that lost.
- **A RabbitMQ subscription whose broker is already gone stops without a warning.** When a stack stops
its broker before its hosts, a stopping subscription's cancel met a connection that had already
closed, or that closed while the cancel waited for its reply, and every subscription of every host
logged `108_107` at Warning — new in 4.4.0, which made a subscription cancel its consumer before it
closes. A closed connection has no consumer left to cancel, which is where the stop was going, so
this is now logged at Debug as `108_007` (`LogEvents.Messaging.SubscriptionAlreadyClosed`). A cancel
that fails while the connection is still up still warns.
- **A handler that completes after the broker went away is no longer reported as failed.** Its
acknowledgement met a closed channel, and the RabbitMQ bus logged `108_101` at Error although the
handler had succeeded, then tried to reject the message on the same closed channel. It now logs
`108_008` (`LogEvents.Messaging.MessageNotSettledChannelClosed`) at Information: the message could
not be settled, and the broker delivers it again.