"Socket Hang Up" and unfinished steps/tests when executing in Docker

[robbiewareham]

We have configured a set of playwright/stencil tests to write results to ReportPortal using the agent-js-playwright and this is working great when trying to execute from our local windows laptops.

However, when we start using in a CI process using Docker to execute (both in Jenkins and debugging locally), we see an intermittent failure to complete steps & tests or tests marked as failed, but we can see in the playwright output that the tests were all completed successfully.

Although the number of times it occurs within a run is small (e.g 1 issue in 77 tests), it is happening quite regularly at a launch level.

• The test that fails itself is not consistent
• We have tried both V1 & V2 endpoints
• We are having to set a proxy to allow the container to contact the RP instance (which is self hosted)
• This happens when both running the container locally and in Jenkins
• We reduced the likelyhood by reducing the number of steps generates BUT we like the verbosity of seeing playwright steps in RP
• We see errors playwright execution generated by the agent-js-playwright

15:25:23  @mytest: URL: http://my.local.reportportal.com:8080/api/v2/myProject/item/ceaae1bf-de48-4fqb-97d1-e2fc5df0c090
15:25:23  @mytests: method: PUT
15:25:23  @mytest:     at /home/jenkins/workspace/workspace/node_modules/@reportportal/agent-js-playwright/node_modules/@reportportal/client-javascript/lib/rest.js:56:15
15:25:23  @mytest:     at processTicksAndRejections (node:internal/process/task_queues:95:5)
15:25:23  @mytest: Failed to finish nested step. Error: socket hang up

My feeling is that the requested isn't reaching the RP server and the hang up is something within docker but wondered is anyone else experienced this issue and hopefully solved it?

[VadimNastoyashchy]

@AmsterGet I have a similar issue, too.

URL: ***/***/item/5a52da6b-56cc-4d32-abe1-d4f96820ad57
method: POST
    at /home/runner/work/vadimnastoyashchy.github.io/vadimnastoyashchy.github.io/e2e/node_modules/@reportportal/client-javascript/lib/rest.js:56:15
    at processTicksAndRejections (node:internal/process/task_queues:105:5)
Failed to start nested step. Error: socket hang up

Github action run: https://github.com/VadimNastoyashchy/vadimnastoyashchy.github.io/actions/runs/17482771065/job/49656288987

Github action run logs:
1_smoke_tests.txt

[AmsterGet]

Hello guys!
@robbiewareham , @VadimNastoyashchy
Please check with the latest version 5.2.3.

[VadimNastoyashchy]

Hello guys! @robbiewareham , @VadimNastoyashchy Please check with the latest version 5.2.3.

@AmsterGet Hi. New version has fixed the issue, but now I receive another one:

Error: timeout of 30000ms exceeded
Failed to start nested step. Error: timeout of 30000ms exceeded
URL: ***/***/item/3ee445c8-522e-4b46-8b3a-5ac3da587180
URL: ***/***/item/3ee445c8-522e-4b46-8b3a-5ac3da587180
method: POST

[AmsterGet]

@VadimNastoyashchy do you use the async api/v2 endpoint?
This allows to support the higher throughput.

In addition to that the timeout and http request retries can be adjusted based on your load and RP instance shape via restClientConfig option.

[VadimNastoyashchy]

@VadimNastoyashchy do you use the async api/v2 endpoint? This allows to support the higher throughput.

In addition to that the timeout and http request retries can be adjusted based on your load and RP instance shape via restClientConfig option.

@AmsterGet Thanks regarding api/v2. My bad, have to use api/v1

[AmsterGet]

@VadimNastoyashchy Perhaps I misunderstood you.
You were using api/v2 and decided to switch to api/v1?

[FariAmrid]

@AmsterGet I am encountering ETIMEDOUT and other errors specifically when using v2 with includeTestSteps: true in a CI environment.

The Symptom:
The test suite runs perfectly locally, but when executed in CI (GitHub Actions), requests fail mid-run. This results in missing test executions in the dashboard (e.g., only 64 out of 275 tests are recorded).

If I set includeTestSteps: false, the errors disappear and stability returns, but I lose the step-level granularity.

Environment:

• Host: AWS EC2 t3.xlarge (4 vCPU, 16GB RAM)
• Deployment: Docker Compose
• Client: Playwright Agent (latest)

Troubleshooting Steps Attempted (None Resolved the Issue):
I have aggressively tuned the infrastructure to rule out resource bottlenecks, but the issue persists only in CI:

1. Network MTU: Fixed Docker network MTU to 1350 (confirmed com.docker.network.driver.mtu: 1350) to prevent packet drops.
2. Resource Upgrade: Upgraded EC2 from t2.large (8GB) to t3.xlarge (16GB RAM).
3. Service Tuning: Increased API Service heap size (JAVA_OPTS) from 1GB to 4GB.
4. Disabled Analyzer: Completely stopped the analyzer service to reduce RabbitMQ load.
5. RabbitMQ Maintenance: Wiped RabbitMQ volumes to clear any stuck/zombie queues.
6. Client Config: Increased restClientConfig.timeout to 5 minutes and set retry: 0 to prevent "retry storms."

Logs:
The failures consistently manifest as connection timeouts from the client side during heavy load (nested steps):

method: POST
at /home/runner/work/playwright/playwright/node_modules/@reportportal/client-javascript/lib/rest.js:93:15
at processTicksAndRejections (node:internal/process/task_queues:105:5)
Error: connect ETIMEDOUT {ip}
URL: http://url/api/v2/cms/item/f6740a0a-db2c-42b8-82a2-ef1b21123db6


method: PUT
    at /home/runner/work/playwright/playwright/node_modules/@reportportal/client-javascript/lib/rest.js:93:15
    at processTicksAndRejections (node:internal/process/task_queues:105:5)
Error: connect ETIMEDOUT {ip}
URL: http://url/api/v1/cms/item/47939f55-89fc-4be9-b057-b3cb6695710d

method: POST
at /home/runner/work/playwright/playwright/node_modules/@reportportal/client-javascript/lib/rest.js:93:15
at processTicksAndRejections (node:internal/process/task_queues:105:5)
Failed to start nested step. Error: timeout of 60000ms exceeded
URL: http://url/api/v2/cms/item/1ac47bbf-dde2-410d-9481-c8c0e44b4fb2

method: POST
    at /home/runner/work/playwright/playwright/node_modules/@reportportal/client-javascript/lib/rest.js:93:15
    at processTicksAndRejections (node:internal/process/task_queues:105:5)
    at runNextTicks (node:internal/process/task_queues:69:3)
    at listOnTimeout (node:internal/timers:555:9)
    at processTimers (node:internal/timers:529:7)
Failed to finish nested step. Error: timeout of 60000ms exceeded
URL: http://url/api/v2/cms/item/0d446757-460a-4af4-b1bd-c0b6e34e9c49

method: PUT
    at /home/runner/work/playwright/playwright/node_modules/@reportportal/client-javascript/lib/rest.js:93:15
    at processTicksAndRejections (node:internal/process/task_queues:105:5)
Failed to finish test. Error: timeout of 60000ms exceeded
URL: http://url/api/v2/cms/item/df72ddc3-93c2-46f1-b624-ecdfcc7c7cae

It appears that the sheer volume of HTTP requests generated by includeTestSteps: true from a high-throughput CI runner overwhelms the API/RabbitMQ ingestion rate, causing it to stop accepting TCP connections. I am happy to test any further recommendations or provide specific logs.