publisher stops suddenly. (deadlock?)

[caoyang1024]

Observed behavior

try to stress test nats, by sending 1000 msg/s each thread, with 1 - 100 different number of threads, when threads go higher, this situation happens more often.

Expected behavior

looks to me like a deadlock, i can provide more info if needed.

Server and client version

"version": "2.10.21"

Host environment

No response

Steps to reproduce

No response

[mtmk]

thanks for the report @caoyang1024 I'm suspecting potentially using semaphore might be causing this and thinking about switching to using a channel instead, but just speculating at the moment. If you could post a minimal console app program.cs and the server setup that would help a lot.

[caoyang1024]

@mtmk sure. here is my testing code. https://github.com/caoyang1024/stress-test.git

i'm running my NATS using docker locally.

services:
  nats:
    image: nats
    ports:
      - "4222:4222"
      - "8222:8222"
    command: "--cluster_name NATS --cluster nats://0.0.0.0:6222 --http_port 8222 "
    networks: ["nats"]
  nats-1:
    image: nats
    ports:
      - "4223:4222"
    command: "--cluster_name NATS --cluster nats://0.0.0.0:6222 --routes=nats://ruser:T0pS3cr3t@nats:6222"
    networks: ["nats"]
    depends_on: ["nats"]
  nats-2:
    image: nats
    ports:
      - "4224:4222"
    command: "--cluster_name NATS --cluster nats://0.0.0.0:6222 --routes=nats://ruser:T0pS3cr3t@nats:6222"
    networks: ["nats"]
    depends_on: ["nats"]

networks:
  nats:
    name: nats
    external: true

[mtmk]

thanks @caoyang1024 unfortunately I wasn't able to reproduce the issue but I seen similar reports before. Are you able to give this branch a go?

• [EXPERIMENT] Channel lock experiment #646