[javier-aliaga]

Description

When a caller cancels an actor call while the method runs, daprd releases the actor and sends the next call for the same actor ID. The SDK kept running the cancelled method, so two calls ran on the same actor at once. Over HTTP the cancelled call also skipped onPostActorMethod, so its state changes were not saved.

ActorManager now runs activation, method calls (timers and reminders included) and deactivation through a new per-actor ActorTurnLock. Calls for one actor run one at a time, in arrival order. Calls for different actors still run in parallel.

• A call cancelled while it runs keeps the actor until the method and onPostActorMethod finish. Its caller stops waiting right away.
• A call cancelled while it waits is removed from the queue and never runs.
• A queued call starts on Schedulers.boundedElastic(), so a long queue does not grow the stack.
• Each call sees the Reactor context of the subscriber that queued it.

Behavior changes:

• Concurrent first calls to the same actor now create one instance. Before, each call could create its own.
• The lock has no timeout and no reentrancy, and the SDK does not configure reentrancy today. A method that never returns blocks every later call to that actor ID. An actor that calls itself waits until the caller's timeout, which is what daprd already does without reentrancy.

Tests:

• ActorTurnLockTest covers ordering, cancellation, errors, empty results, a 10,000-call queue and context propagation.
• ActorManagerTest adds three cases. A cancelled call still runs onPostActorMethod before the next call starts. Ten concurrent activations create one actor. Deactivation waits for the running call.
• ActorCancelledCallIT (HTTP app) and ActorCancelledCallGrpcHostIT (actor event stream) run against daprd. Each cancels five 2-second calls after 500 ms and checks that only one ran at a time. Both fail on master with 4 calls running at once, and pass with this change.

Issue reference

Fixes #1871

Checklist

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

[Copilot]

Copilot was unable to review this pull request because the user who requested the review has insufficient remaining quota.

[codecov]

Codecov Report

✅ All modified and coverable lines are covered by tests.
✅ Project coverage is 80.63%. Comparing base (00f769f) to head (a4e1158).

Additional details and impacted files

@@             Coverage Diff              @@
##             master    #1877      +/-   ##
============================================
+ Coverage     80.52%   80.63%   +0.10%     
- Complexity     2835     2850      +15     
============================================
  Files           274      275       +1     
  Lines          8412     8459      +47     
  Branches        905      911       +6     
============================================
+ Hits           6774     6821      +47     
  Misses         1272     1272              
  Partials        366      366              

☔ View full report in Codecov by Harness.
📢 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.