fix: proactor_threads is respected in k8s

- Honor --proactor_threads flag over cgroup-derived CPU limit
- Log when flag overrides cgroup value (helps debugging in production)
- Add CI job to verify behavior under real Docker cgroup limits

Author: starek4

.github/workflows/ci.yml

@@ -187,6 +187,59 @@ jobs:
           name: regression_logs
           path: /tmp/failed/*
 
+      - name: Upload dragonfly binary for cgroup test
+        if: matrix.container == 'ubuntu-dev:24' && matrix.build-type == 'Release' && matrix.sanitizers == 'NoSanitizers' && matrix.compiler.cxx == 'g++'
+        uses: actions/upload-artifact@v7
+        with:
+          name: dragonfly-binary
+          path: ${{ github.workspace }}/build/dragonfly
+          retention-days: 1
+
+  cgroup-thread-detection:
+    runs-on: ubuntu-latest
+    needs: [build]
+    steps:
+      - uses: actions/checkout@v6
+
+      - uses: actions/download-artifact@v7
+        with:
+          name: dragonfly-binary
+          path: build
+
+      - name: Test cgroup CPU auto-detection
+        run: |
+          chmod +x build/dragonfly
+
+          # Scenario 1: --cpus=2, no flag -> should auto-detect 2 threads
+          CID=$(docker run --rm -d --cpus=2 --security-opt seccomp=unconfined \
+            -v "$PWD/build/dragonfly:/dragonfly" -p 16380:6379 ubuntu:24.04 \
+            /dragonfly --port 6379 --maxmemory 1G --dbfilename "" --noversion_check)
+
+          for i in $(seq 1 20); do
+            sleep 1
+            redis-cli -p 16380 PING 2>/dev/null | grep -q PONG && break
+          done
+
+          THREADS=$(redis-cli -p 16380 INFO server | grep -oP 'thread_count:\K\d+')
+          docker rm -f "$CID"
+          echo "Scenario 1: expected 2, got $THREADS"
+          [ "$THREADS" -eq 2 ]
+
+          # Scenario 2: --cpus=2 + --proactor_threads=4 -> flag must win
+          CID=$(docker run --rm -d --cpus=2 --security-opt seccomp=unconfined \
+            -v "$PWD/build/dragonfly:/dragonfly" -p 16381:6379 ubuntu:24.04 \
+            /dragonfly --port 6379 --proactor_threads 4 --maxmemory 1G --dbfilename "" --noversion_check)
+
+          for i in $(seq 1 20); do
+            sleep 1
+            redis-cli -p 16381 PING 2>/dev/null | grep -q PONG && break
+          done
+
+          THREADS=$(redis-cli -p 16381 INFO server | grep -oP 'thread_count:\K\d+')
+          docker rm -f "$CID"
+          echo "Scenario 2: expected 4, got $THREADS"
+          [ "$THREADS" -eq 4 ]
+
   lint-test-chart:
     runs-on: ubuntu-latest
     needs: [build]

src/server/dfly_main.cc

@@ -82,6 +82,7 @@ ABSL_DECLARE_FLAG(uint32_t, memcached_port);
 ABSL_DECLARE_FLAG(uint16_t, admin_port);
 ABSL_DECLARE_FLAG(std::string, admin_bind);
 ABSL_DECLARE_FLAG(strings::MemoryBytesFlag, maxmemory);
+ABSL_DECLARE_FLAG(uint32_t, proactor_threads);
 
 ABSL_FLAG(string, bind, "",
           "Bind address. If empty - binds on all interfaces. "
@@ -1080,6 +1081,14 @@ Usage: dragonfly [FLAGS]
 
 #ifdef __linux__
   UpdateResourceLimitsIfInsideContainer(&mem_info, &max_available_threads);
+  // If --proactor_threads (or DFLY_proactor_threads env var) was explicitly set by the user,
+  // honor it over the cgroup-derived CPU limit. The flag defaults to 0, so any non-zero value
+  // means the user explicitly requested a specific thread count.
+  if (absl::GetFlag(FLAGS_proactor_threads) > 0) {
+    LOG(INFO) << "Using proactor_threads=" << absl::GetFlag(FLAGS_proactor_threads)
+              << " (overriding cgroup-derived " << max_available_threads << ")";
+    max_available_threads = 0;  // causes ProactorPool to use FLAGS_proactor_threads
+  }
 #endif
 
   if (mem_info.swap_total != 0)

tests/dragonfly/proactor_threads_test.py

@@ -0,0 +1,66 @@
+"""
+Tests that --proactor_threads is respected even inside containers with CPU limits.
+
+Background: UpdateResourceLimitsIfInsideContainer() reads the cgroup CPU quota and
+passes it as pool_size to ProactorPool. When pool_size != 0, helio's ProactorPool
+ignores FLAGS_proactor_threads entirely. The fix resets pool_size to 0 when
+--proactor_threads (or DFLY_proactor_threads) is explicitly set, so the flag wins.
+"""
+
+import pytest
+from .instance import DflyInstanceFactory
+
+
+@pytest.mark.asyncio
+async def test_proactor_threads_flag_is_respected(df_factory: DflyInstanceFactory):
+    """Starting with --proactor_threads=N must result in exactly N threads."""
+    server = df_factory.create(proactor_threads=3)
+    server.start()
+    client = server.client()
+    try:
+        info = await client.info("server")
+        assert int(info["thread_count"]) == 3
+    finally:
+        await client.aclose()
+        server.stop()
+
+
+@pytest.mark.asyncio
+async def test_proactor_threads_env_var_is_respected(df_factory: DflyInstanceFactory, monkeypatch):
+    """DFLY_proactor_threads env var must behave identically to the CLI flag.
+
+    The subprocess inherits the parent environment, so setting the variable here
+    is equivalent to setting it in a Kubernetes pod's env: section.
+    """
+    monkeypatch.setenv("DFLY_proactor_threads", "2")
+    # No proactor_threads kwarg — only the env var should drive the thread count.
+    server = df_factory.create()
+    server.start()
+    client = server.client()
+    try:
+        info = await client.info("server")
+        assert int(info["thread_count"]) == 2
+    finally:
+        await client.aclose()
+        server.stop()
+
+
+@pytest.mark.asyncio
+async def test_proactor_threads_flag_overrides_env_var(
+    df_factory: DflyInstanceFactory, monkeypatch
+):
+    """CLI flag takes priority over DFLY_proactor_threads env var.
+
+    ParseFlagsFromEnv() skips env vars when the flag was already set on the
+    command line (WasPresentOnCommandLine check), so the CLI value must win.
+    """
+    monkeypatch.setenv("DFLY_proactor_threads", "2")
+    server = df_factory.create(proactor_threads=4)
+    server.start()
+    client = server.client()
+    try:
+        info = await client.info("server")
+        assert int(info["thread_count"]) == 4
+    finally:
+        await client.aclose()
+        server.stop()