Gracefully shutdown app #1543
Open
Gracefully shutdown app #1543
+82
−0
Conversation
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
martinothamar
added
other
📝 WalkthroughWalkthroughThe changes introduce graceful shutdown handling for the Altinn.App.Api application. A new Changes
Estimated code review effort🎯 3 (Moderate) | ⏱️ ~20 minutes
Pre-merge checks and finishing touches❌ Failed checks (1 warning)
✅ Passed checks (2 passed)
✨ Finishing touches
🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
|
/publish |
PR release:
|
![github-advanced-security[bot]](https://avatars.githubusercontent.com/in/57789?s=60&v=4){kind=small}
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cs
Dismissed
Show dismissed
Hide dismissed
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cs
Dismissed
Show dismissed
Hide dismissed
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cs
Dismissed
Show dismissed
Hide dismissed
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🧹 Nitpick comments (3)
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cs (3)
1-1: Remove unused using
System.Runtime.InteropServicesisn’t used.-using System.Runtime.InteropServices;
18-30: Add OS guard to avoid PlatformNotSupported on WindowsOnly register POSIX signals on non-Windows OS. Keeps dev/test on Windows safe.
- public Task WaitForStartAsync(CancellationToken cancellationToken) + public Task WaitForStartAsync(CancellationToken cancellationToken) { - if (!_environment.IsDevelopment()) + if (!_environment.IsDevelopment() && !OperatingSystem.IsWindows()) { _disposables = [ PosixSignalRegistration.Create(PosixSignal.SIGINT, HandleSignal), PosixSignalRegistration.Create(PosixSignal.SIGQUIT, HandleSignal), PosixSignalRegistration.Create(PosixSignal.SIGTERM, HandleSignal), ]; } return Task.CompletedTask; }
32-42: Prevent multiple StopApplication calls and prefer async/await over ContinueWithMake shutdown trigger idempotent and avoid ContinueWith; run the delayed stop on the thread pool.
- private void HandleSignal(PosixSignalContext ctx) + private void HandleSignal(PosixSignalContext ctx) { _logger.LogInformation("Received shutdown signal: {Signal}, delaying shutdown", ctx.Signal); ctx.Cancel = true; // Signal intercepted here, we are now responsible for calling `StopApplication` - Task.Delay(_delay) - .ContinueWith(t => - { - _logger.LogInformation("Starting host shutdown..."); - _applicationLifetime.StopApplication(); - }); + _ = Task.Run(async () => + { + await Task.Delay(_delay).ConfigureAwait(false); + if (Interlocked.Exchange(ref _stopRequested, 1) == 0) + { + _logger.LogInformation("Starting host shutdown..."); + _applicationLifetime.StopApplication(); + } + }); }Add this field near the top of the class:
private int _stopRequested;
📜 Review details
Configuration used: CodeRabbit UI
Review profile: CHILL
Plan: Pro
📒 Files selected for processing (2)
src/Altinn.App.Api/Extensions/ServiceCollectionExtensions.cs(3 hunks)src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cs(1 hunks)
🧰 Additional context used
📓 Path-based instructions (3)
**/*.cs
📄 CodeRabbit inference engine (CLAUDE.md)
**/*.cs: Use internal accessibility on types by default
Use sealed for classes unless inheritance is a valid use-case
Dispose IDisposable/IAsyncDisposable instances
Do not use .GetAwaiter().GetResult(), .Result(), .Wait(), or other blocking APIs on Task
Do not use the Async suffix for async methods
Write efficient code; avoid unnecessary allocations (e.g., avoid repeated ToString calls; consider for-loops over LINQ when appropriate)
Do not invoke the same async operation multiple times in the same code path unless necessary
Avoid awaiting async operations inside tight loops; prefer batching with a sensible upper bound on parallelism
Use CSharpier for formatting (required before commits; formatting also runs on build via CSharpier.MSBuild)
Files:
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cssrc/Altinn.App.Api/Extensions/ServiceCollectionExtensions.cs
**/*.{cs,csproj}
📄 CodeRabbit inference engine (CLAUDE.md)
Use Nullable Reference Types
Files:
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cssrc/Altinn.App.Api/Extensions/ServiceCollectionExtensions.cs
src/{Altinn.App.Api,Altinn.App.Core}/**/*.cs
📄 CodeRabbit inference engine (CLAUDE.md)
src/{Altinn.App.Api,Altinn.App.Core}/**/*.cs: Types meant to be implemented by apps should be marked with the ImplementableByApps attribute
For HTTP APIs, define DTOs with ...Request and ...Response naming
Files:
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cssrc/Altinn.App.Api/Extensions/ServiceCollectionExtensions.cs
🧬 Code graph analysis (1)
src/Altinn.App.Api/Extensions/ServiceCollectionExtensions.cs (1)
src/Altinn.App.Api/Infrastructure/Lifetime/AppHostLifetime.cs (1)
AppHostLifetime(7-49)
⏰ Context from checks skipped due to timeout of 90000ms. You can increase the timeout in your CodeRabbit configuration to a maximum of 15 minutes (900000ms). (5)
- GitHub Check: Analyze (csharp)
- GitHub Check: Run dotnet build and test (macos-latest)
- GitHub Check: Run dotnet build and test (ubuntu-latest)
- GitHub Check: Static code analysis
- GitHub Check: Run dotnet build and test (windows-latest)
🔇 Additional comments (1)
src/Altinn.App.Api/Extensions/ServiceCollectionExtensions.cs (1)
9-9: LGTM on namespace importImport is correct and scoped to the new lifetime type.
Comment on lines
+128
to
+129
| ConfigureGracefulShutdown(services); | ||
|
|
There was a problem hiding this comment.
Do not override ConsoleLifetime in Development; gate registration to tt02/prod
Registering a custom IHostLifetime unconditionally replaces ConsoleLifetime, which can break Ctrl+C handling locally. Limit registration to staging++ (tt02) and prod, and pass env to the helper.
Apply:
- ConfigureGracefulShutdown(services);
+ ConfigureGracefulShutdown(services, env);And update the helper signature and gating (see diff on the method below).
Committable suggestion skipped: line range outside the PR's diff.
🤖 Prompt for AI Agents
In src/Altinn.App.Api/Extensions/ServiceCollectionExtensions.cs around lines
128-129, ConfigureGracefulShutdown(services) is registering a custom
IHostLifetime unconditionally which overrides ConsoleLifetime in Development;
change the call to accept the current environment and only register the custom
lifetime when the environment is tt02 or Production. Update the
ConfigureGracefulShutdown helper signature to accept IHostEnvironment (or
environment name) and move the conditional there so that in Development the
default ConsoleLifetime remains untouched while in tt02/prod the custom
IHostLifetime is added to the services.
Comment on lines
+552
to
+581
|
|
||
| private static void ConfigureGracefulShutdown(IServiceCollection services) | ||
| { | ||
| // Need to coordinate graceful shutdown (let's assume k8s as the scheduler/runtime): | ||
| // - deployment is configured with a terminationGracePeriod of 30s (default timeout before SIGKILL) | ||
| // - k8s flow of information is eventually consistent. | ||
| // it takes time for knowledge of SIGTERM on the worker node to propagate to e.g. networking layers | ||
| // (k8s Service -> Endspoints rotation. It takes time to be taken out of Endpoint rotation) | ||
| // - we want to gracefully drain ASP.NET core for requests, leaving some time for active requests to complete | ||
| // This leaves us with the following sequence of events | ||
| // - container receives SIGTERM | ||
| // - `AppHostLifetime` intercepts SIGTERM and delays for `shutdownDelay` | ||
| // - `AppHostLifetime` calls `IHostApplicationLifetime.StopApplication`, to start ASP.NET Core shutdown process | ||
| // - ASP.NET Core will spend a maximum of `shutdownTimeout` trying to drain active requests | ||
| // (cancelable requests can combine cancellation tokens with `IHostApplicationLifetime.ApplicationStopping`) | ||
| // - If ASP.NET Core completes shutdown within `shutdownTimeout`, everything is fine | ||
| // - If ASP.NET Core is stuck or in some way can't terminate, kubelet will eventually SIGKILL | ||
| var shutdownDelay = TimeSpan.FromSeconds(5); | ||
| var shutdownTimeout = TimeSpan.FromSeconds(20); | ||
|
|
||
| services.AddSingleton<IHostLifetime>(serviceProvider => | ||
| { | ||
| var logger = serviceProvider.GetRequiredService<ILogger<AppHostLifetime>>(); | ||
| var env = serviceProvider.GetRequiredService<IHostEnvironment>(); | ||
| var lifetime = serviceProvider.GetRequiredService<IHostApplicationLifetime>(); | ||
| return new AppHostLifetime(logger, env, lifetime, shutdownDelay); | ||
| }); | ||
|
|
||
| services.Configure<HostOptions>(options => options.ShutdownTimeout = shutdownTimeout); | ||
| } |
There was a problem hiding this comment.
Gate by environment; make idempotent values configurable; fix typo
- Gate lifetime registration to staging++ (tt02) and Production only.
- Consider configuring delay/timeout from config.
- Minor: “Endspoints” -> “Endpoints”.
- private static void ConfigureGracefulShutdown(IServiceCollection services)
+ private static void ConfigureGracefulShutdown(IServiceCollection services, IHostEnvironment env)
{
- // Need to coordinate graceful shutdown (let's assume k8s as the scheduler/runtime):
+ // Need to coordinate graceful shutdown (let's assume k8s as the scheduler/runtime):
// - deployment is configured with a terminationGracePeriod of 30s (default timeout before SIGKILL)
// - k8s flow of information is eventually consistent.
- // it takes time for knowledge of SIGTERM on the worker node to propagate to e.g. networking layers
- // (k8s Service -> Endspoints rotation. It takes time to be taken out of Endpoint rotation)
+ // it takes time for knowledge of SIGTERM on the worker node to propagate to e.g. networking layers
+ // (k8s Service -> Endpoints rotation. It takes time to be taken out of Endpoint rotation)
// - we want to gracefully drain ASP.NET core for requests, leaving some time for active requests to complete
// This leaves us with the following sequence of events
// - container receives SIGTERM
// - `AppHostLifetime` intercepts SIGTERM and delays for `shutdownDelay`
// - `AppHostLifetime` calls `IHostApplicationLifetime.StopApplication`, to start ASP.NET Core shutdown process
// - ASP.NET Core will spend a maximum of `shutdownTimeout` trying to drain active requests
// (cancelable requests can combine cancellation tokens with `IHostApplicationLifetime.ApplicationStopping`)
// - If ASP.NET Core completes shutdown within `shutdownTimeout`, everything is fine
// - If ASP.NET Core is stuck or in some way can't terminate, kubelet will eventually SIGKILL
- var shutdownDelay = TimeSpan.FromSeconds(5);
- var shutdownTimeout = TimeSpan.FromSeconds(20);
+ var shutdownDelay = TimeSpan.FromSeconds(
+ services.BuildServiceProvider().GetRequiredService<IConfiguration>()
+ .GetValue("GracefulShutdown:DelaySeconds", 5)
+ );
+ var shutdownTimeout = TimeSpan.FromSeconds(
+ services.BuildServiceProvider().GetRequiredService<IConfiguration>()
+ .GetValue("GracefulShutdown:TimeoutSeconds", 20)
+ );
- services.AddSingleton<IHostLifetime>(serviceProvider =>
- {
- var logger = serviceProvider.GetRequiredService<ILogger<AppHostLifetime>>();
- var env = serviceProvider.GetRequiredService<IHostEnvironment>();
- var lifetime = serviceProvider.GetRequiredService<IHostApplicationLifetime>();
- return new AppHostLifetime(logger, env, lifetime, shutdownDelay);
- });
+ // Only wire up in staging++ (tt02) and Production to avoid replacing ConsoleLifetime in Development/test-like envs
+ if (IsStagingPlus(env))
+ {
+ services.AddSingleton<IHostLifetime>(serviceProvider =>
+ {
+ var logger = serviceProvider.GetRequiredService<ILogger<AppHostLifetime>>();
+ var lifetime = serviceProvider.GetRequiredService<IHostApplicationLifetime>();
+ return new AppHostLifetime(logger, env, lifetime, shutdownDelay);
+ });
+ }
services.Configure<HostOptions>(options => options.ShutdownTimeout = shutdownTimeout);
}
+
+ private static bool IsStagingPlus(IHostEnvironment env) =>
+ env.IsProduction() || string.Equals(env.EnvironmentName, "tt02", StringComparison.OrdinalIgnoreCase);Committable suggestion skipped: line range outside the PR's diff.
|
bjorntore
approved these changes
Oct 29, 2025
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Description
I think we want to gracefull shutdown in Staging++ (which means tt02 and prod mostly).
If we want to have process/next locks: Altinn/altinn-studio#16669
this means we can make it more unlikely to hit timeout/cancellation-related cases during rolling upgrades and such (let process/next complete and release any locks held in Storage)
Related Issue(s)
Verification
Documentation
Summary by CodeRabbit