bug: concurrent running of ElasticSearch tests may produce interferences due to same ports in use

[ghost]

This is a speculative resolution for a problem we have noticed only recently.

One (or more) of the integration tests for ElasticSearch was failing to reindex a model when during the build on the Jenkins side, showing a socket timeout exception.

Given how those tests work (an ES runtime is launched by Allegro on specific ports), I suspect that multiple concurrent builds testing the ES integration tests at the same time might actually be stepping on each others' feet, when the build processes are not isolated.

I will try and randomize the ports used to spin the service with the following rules:

• must be 4 digit ports
• must not be the production ES ports, or other obvious entries

Hopefully this will limit the collisions well enough for the use case.

[ghost]

PR here.
Local tests doing good - hopefully no surprise upstream.

[ghost]

Looks like this solved one problem, aka the socket timeouts, but still does not allow multiple non-isolated builds running the ElasticSearch tests concurrently - maybe it's just a limitation on Allegro's side...
Working with @t-gauss to see if we can't force throttling builds.

Not sure what could be an alternative solution should that fail, save for disabling those tests (which does not sound like a great idea).

[t-gauss]

Hi @mena-bosch ,
being still a newbie in this topic I guess adding disableConcurrentBuilds() in the options block of the jenkinsfile could help as a temporary workaround.

from the docs:
disableConcurrentBuilds
Disallow concurrent executions of the Pipeline. Can be useful for preventing simultaneous accesses to shared resources, etc. For example: options { disableConcurrentBuilds() }

[t-gauss]

PR here.

[ghost]

Thanks @t-gauss . This one was merged by @kolotu now so closing.