Dapr.Testcontainers: probe gRPC port with HTTP/2 preface to eliminate transient Unavailable on first RPC

[Copilot]

Description

Integration tests intermittently fail on the first RPC with:

Grpc.Core.RpcException: Status(StatusCode="Unavailable",
  Detail="Error connecting to subchannel. ... Connection refused")

Root cause: DaprdContainer.StartAsync only validates the gRPC port at the TCP layer. A successful TCP connect proves Docker's port forwarding accepts SYNs — it does not prove the upstream HTTP/2 (gRPC) server is actively serving. On slow CI hosts there is a brief window between Docker publishing the port and daprd's gRPC Serve() loop being wired up; the test's first RPC lands in that window and gets Unavailable. The HTTP reachability probe already in place only covers the HTTP server on a different port.

For reference, daprd's startup log order (dapr/dapr/pkg/runtime/runtime.go L790/L802/L813) is API gRPC → HTTP → Internal gRPC, so the existing "Internal gRPC server is running" wait is in fact the last of the three and is not the source of the flake. An earlier commit on this branch that changed that wait message is reverted here.

Changes:

• ContainerReadinessProbe.WaitForGrpcServerReadyAsync(host, port, timeout, ct) — new probe that performs the HTTP/2 connection handshake itself: opens TCP, sends the 24-byte client preface (RFC 7540 §3.5: PRI * HTTP/2.0\r\n\r\nSM\r\n\r\n), and waits for any bytes back from the server's SETTINGS frame. Per-attempt bound of 2 s, 200 ms inter-attempt backoff, retries SocketException/IOException/attempt-timeout until the overall timeout. Dependency-free; works against any HTTP/2 (gRPC) server.
• DaprdContainer.StartAsync — invokes the new probe against 127.0.0.1:GrpcPort as the final readiness gate, alongside the existing HTTP reachability probe for the HTTP port.
• ContainerReadinessProbeTests — 4 tests: HTTP/2 server preface → returns; accepts-but-never-responds (simulates the CI race) → TimeoutException; port refused → retries then times out; pre-cancelled token → OperationCanceledException.
• Reverts the earlier wait-message change in DaprdContainer.cs; the log line is back to "Internal gRPC server is running".

Out of scope:

• test/Dapr.E2E.Test/DaprTestApp.cs (uses the dapr run CLI, different startup mechanism).
• PortUtilities.GetAvailablePort releases the socket before returning the number — TOCTOU race used in app-first mode, already wrapped in DaprTestApplicationBuilder's retry loop.

Issue reference

Checklist

• Code compiles correctly
• Created/updated tests
• Extended the documentation

[codecov]

Codecov Report

❌ Patch coverage is 87.80488% with 5 lines in your changes missing coverage. Please review.
✅ Project coverage is 65.01%. Comparing base (62d3a6e) to head (cc369fa).

Files with missing lines	Patch %	Lines
...r.Testcontainers/Common/ContainerReadinessProbe.cs	86.11%	3 Missing and 2 partials ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##           master    #1821      +/-   ##
==========================================
+ Coverage   64.90%   65.01%   +0.10%     
==========================================
  Files         300      300              
  Lines        8833     8874      +41     
  Branches     1040     1043       +3     
==========================================
+ Hits         5733     5769      +36     
- Misses       2846     2849       +3     
- Partials      254      256       +2     

Flag	Coverage Δ
net10.0	64.99% <87.80%> (+0.10%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.