[Fleet] Improve performance of Fleet preconfiguration integration tests

[juliaElastic]

Fleet has integration tests that start up an Elasticsearch instance: https://github.com/elastic/kibana/blob/main/x-pack/plugins/fleet/server/integration_tests/reset_preconfiguration.test.ts#L166

These are slow to run on CI, and also slow to run locally to test changes.

Can we improve these by using a common ES instance (that would speed up CI)?

For local development, it would be great to have a way to separately start an ES instance and keep it running while running the tests multiple times (and inspect the contents of indices).

FTR already has these features, is it possible to rewrite these integration tests to run by FTR?

PASS  x-pack/plugins/fleet/server/integration_tests/ha_setup.test.ts (276.963 s)
PASS  x-pack/plugins/fleet/server/integration_tests/upgrade_package_install_version.test.ts (103.607 s)
PASS  x-pack/plugins/fleet/server/integration_tests/upgrade_agent_policy_schema_version.test.ts (104.973 s)
PASS  x-pack/plugins/fleet/server/integration_tests/fleet_usage_telemetry.test.ts (79.585 s)

This test is skipped: https://github.com/elastic/kibana/blob/main/x-pack/plugins/fleet/server/integration_tests/cloud_preconfiguration.test.ts#L26

[elasticmachine]

Pinging @elastic/fleet (Team:Fleet)

[nchaulet]

@juliaElastic @kpollich I am not sure if we should move those tests to FTR, and looking at this it seems hard to do so for most of them, as we use specific kibana configuration and some jest feature. I think we should keep the number of these tests to the minimum but they seems very valuable as they start with some fresh and clean ES and do not have side effects.

For local development, it would be great to have a way to separately start an ES instance and keep it running while running the tests multiple times (and inspect the contents of indices).

We can inspect the content of the indices of this test by using node debugger while running the test, I can add some instructions on how to do this.

[nchaulet]

I think we can close that issue for now, feel free to reopen it if you see anything that can be improved