[Ingest Manager] prevent possible race conditions when installing/update packages from multiple requests

[neptunian]

Now that we are tracking the installation/update process in the saved object, we should not try to install/update a package if its currently installing. For example users from different kibana instances trying to update a package or accidentally posting directly to the endpoint twice. This could cause some unexpected behaviour and it would be good to guard against this scenario. We could check the status and return a message that the package is already currently installing or updating. @ruflin mentioned that it might not be reliable to read from the saved object whether something is installing as it could be delayed after writing. Some research needs to be done to make sure this is a viable solution.

This already exists in the UI where we block the install button while its installing, but it doesnt prevent it from happening with multiple kibana instances or through the api.

Would need to be mindful of the scenario that the package is stuck installing but it was because of a kibana crash, that we keep the ability to reinstall or reupdate.

[elasticmachine]

Pinging @elastic/ingest-management (Team:Ingest Management)

[ph]

race conditions are hard, if the operation are idempotent and update to SO are using optimistic locking with version this should be fine?

[ruflin]

The question is how SO works under the hood. This reminded me also of #75497

[elasticmachine]

Pinging @elastic/endpoint-management (Team:Endpoint Management)

[ph]

Going to mark this as bug.

[skh]

Implemented in #84190

[muskangulati-qasource]

Bug Conversion

• Test-Case not required as this is a Dev Validation ticket

Thanks!