You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
For gRPC-to-gRPC Streaming, when multiple clients hit the gateway only 5 client connections successful and are able to send and receive messages through gateway.The gateway assigns worker threads to first 5 client calls. The other clients send messages to gateway but are not assigned a worker request to proceed with streaming. These clients are waiting till the other client closes connections.
Steps to Reproduce
1)git clone https://github.com/project-flogo/grpc
cd grpc/examples/api/grpc-to-grpc
go build
2)Start proxy gateway:
./grpc-to-grpc
3)Start sample gRPC server.
./grpc-to-grpc -server
4)Run 6 clients in different terminals
./grpc-to-grpc -client -port 9096 -method bulkusers
5)It is seen that the 5 clients are assigned workers but 6th client is not assigned a worker and is kept in wait state until a previous client is killed.
Flogo uses a runner to execute actions. There are two types of runners: DIRECT and POOLED. The default runner type is POOLED. POOLED uses a pool of runners (default pool size is 5) to execute actions. The 'bulkusers' method ties up the workers which prevents other requests from happening. The DIRECT runner directly executes the action and doesn't have a limit on the number of simultaneous actions being executed.
For gRPC-to-gRPC Streaming, when multiple clients hit the gateway only 5 client connections successful and are able to send and receive messages through gateway.The gateway assigns worker threads to first 5 client calls. The other clients send messages to gateway but are not assigned a worker request to proceed with streaming. These clients are waiting till the other client closes connections.
Steps to Reproduce
1)git clone https://github.com/project-flogo/grpc
cd grpc/examples/api/grpc-to-grpc
go build
2)Start proxy gateway:
./grpc-to-grpc
3)Start sample gRPC server.
./grpc-to-grpc -server
4)Run 6 clients in different terminals
./grpc-to-grpc -client -port 9096 -method bulkusers
5)It is seen that the 5 clients are assigned workers but 6th client is not assigned a worker and is kept in wait state until a previous client is killed.
Note:
It is observed that FLOGO_RUNNER_WORKERS default value is 5. The clients connecting is directly proportional to FLOGO_RUNNER_WORKERS value.
Also the same is not observed using the old mashling code.
https://github.com/TIBCOSoftware/mashling/
https://github.com/TIBCOSoftware/mashling-recipes
The text was updated successfully, but these errors were encountered: