Failing test: Serverless Search API Integration Tests - Common Group 1.x-pack/test_serverless/api_integration/test_suites/common/index_management/inference_endpoints·ts - Index Management APIs Inference endpoints "before all" hook in "Inference endpoints"

[kibanamachine]

A test failed on a tracked branch

Error: Expected status code 200, got 500 with body '{"error":{"root_cause":[{"type":"status_exception","reason":"Model importing failed due to an Exception [java.lang.IllegalArgumentException: [definition_length] must be greater than 0]"}],"type":"status_exception","reason":"Model importing failed due to an Exception [java.lang.IllegalArgumentException: [definition_length] must be greater than 0]","caused_by":{"type":"illegal_argument_exception","reason":"[definition_length] must be greater than 0"}},"status":500}'
    at Assertion.assert (expect.js:100:11)
    at Assertion.eql (expect.js:244:8)
    at Object.assertResponseStatusCode (api.ts:92:31)
    at Object.createInferenceEndpoint (api.ts:243:12)
    at processTicksAndRejections (node:internal/process/task_queues:95:5)
    at Context.<anonymous> (inference_endpoints.ts:27:9)
    at Object.apply (wrap_function.js:73:16) {
  actual: '500',
  expected: '200',
  showDiff: true
}

First failure: kibana-on-merge - main

[elasticmachine]

Pinging @elastic/kibana-management (Team:Kibana Management)

[alisonelizabeth]

@saarikabhasi or @saikatsarkar056 this something one of you would be able to take a look at?

[saarikabhasi]

@alisonelizabeth I am not able to reproduce this error locally. Can we close this ticket?

[alisonelizabeth]

If you have verified the test case and the tests are passing again we can close and see if the issue resurfaces. However, I see another similar issue opened against the O11y tests so I wonder if further investigation is warranted - https://github.com/elastic/kibana/issues/184961

[kibanamachine]

New failure: kibana-on-merge - main

[saarikabhasi]

PR to delete underlying endpoint during clean up: https://github.com/elastic/kibana/pull/187521

[saarikabhasi]

Tried to reproduce this issue but unfortunately could not against QA cloud deployment and locally. Added a fix to handle request time out error and delete underlying model as well during clean up.

[saarikabhasi]

Closing this ticket with the https://github.com/elastic/kibana/pull/187521

In the PR:

1. Modified test cases to use kibana internal path instead of directly calling es endpoint
2. clean up underlying trained model
3. un skipped the tests for now to confirm if this helps fixing flaky tests