Skip to content

fix: do not start stream with direct executor #13945

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
mutianf merged 1 commit into from
Jul 29, 2026
+16 −10
Merged

fix: do not start stream with direct executor #13945

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
1 commit
Select commit
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
13 changes: 8 additions & 5 deletions .../src/main/java/com/google/cloud/bigtable/data/v2/internal/channels/ChannelPoolDpImpl.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -26,7 +26,6 @@
import com.google.common.base.Ticker;
import com.google.common.collect.HashMultiset;
import com.google.common.collect.Multiset;
import com.google.common.util.concurrent.MoreExecutors;
import io.grpc.CallOptions;
import io.grpc.ClientCall;
import io.grpc.ManagedChannel;
Expand Down Expand Up @@ -296,11 +295,15 @@ public synchronized SessionStream newStream(
fw.numOutstanding++;
totalStreams++;

// DirectExecutor: gRPC/Netty delivers SessionStream.Listener callbacks directly on the
// I/O thread. All work must be fast and non-blocking; blocking work goes to sessionSyncContext.
// Do NOT override the call executor with directExecutor(): SessionStream.Listener callbacks
// acquire the SessionPool lock (onReady/onGoAway/onClose/onVRpcComplete). directExecutor()
// delivers those callbacks inline on the Netty event-loop (I/O) thread, so a contended pool
// lock blocks the event loop; the blocked event loop then can't deliver the very completions
// that would release sessions and drain the lock -> self-sustaining pod-wide wedge. Leaving
// the executor unset delivers callbacks on gRPC's off-loop channel executor, keeping
// pool-lock acquisition off the transport threads.
ClientCall<SessionRequest, SessionResponse> innerCall =
channelWrapper.channel.newCall(
desc, callOptions.withExecutor(MoreExecutors.directExecutor()));
channelWrapper.channel.newCall(desc, callOptions);

return new SessionStreamImpl(innerCall) {
// null until onBeforeSessionStart fires
Expand Down
13 changes: 8 additions & 5 deletions .../src/main/java/com/google/cloud/bigtable/data/v2/internal/channels/SingleChannelPool.java
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -19,7 +19,6 @@
import com.google.bigtable.v2.SessionClientConfiguration.ChannelPoolConfiguration;
import com.google.bigtable.v2.SessionRequest;
import com.google.bigtable.v2.SessionResponse;
import com.google.common.util.concurrent.MoreExecutors;
import io.grpc.CallOptions;
import io.grpc.ManagedChannel;
import io.grpc.MethodDescriptor;
Expand All @@ -46,10 +45,14 @@ public void close() {
@Override
public SessionStream newStream(
MethodDescriptor<SessionRequest, SessionResponse> desc, CallOptions callOptions) {
// DirectExecutor: gRPC/Netty delivers SessionStream.Listener callbacks directly on the
// I/O thread. All work must be fast and non-blocking; blocking work goes to sessionSyncContext.
return new SessionStreamImpl(
channel.newCall(desc, callOptions.withExecutor(MoreExecutors.directExecutor())));
// Do NOT override the call executor with directExecutor(): SessionStream.Listener callbacks
// acquire the SessionPool lock (onReady/onGoAway/onClose/onVRpcComplete). directExecutor()
// delivers those callbacks inline on the Netty event-loop (I/O) thread, so a contended pool
// lock blocks the event loop; the blocked event loop then can't deliver the very completions
// that would release sessions and drain the lock -> self-sustaining pod-wide wedge. Leaving
// the executor unset delivers callbacks on gRPC's off-loop channel executor, keeping
// pool-lock acquisition off the transport threads.
return new SessionStreamImpl(channel.newCall(desc, callOptions));
}

@Override
Expand Down
Loading