[META] problems surfacing due switching docker base image

[kares]

Logstash is (recently) switched it's base Docker image from CentOS 7 to Ubuntu 20.04

This issue is meant to track (mostly plugin testing) related issues caused by the switch.

• Test: refactor setup and get all tests passing logstash-plugins/logstash-input-beats#442
• ?trying to upgrade rufus-scheduler in http_poller input leads to an issue while interrupting the plugin
  • interrupt example failing on input ES and others -> cause is likely not due rufus-scheduler upgrade
• input exec long sub-process spawn times sample run [Test] Add some extra wiggle room for scheduler to start logstash-plugins/logstash-input-exec#32
• a lot of integration test failures on JMS input, sample [Test] Fix integration tests for 8.x logstash-plugins/logstash-input-jms#53

[kares]

since 94d04f8 this should now be surfacing with 7.x SNAPSHOT builds as well

[kares]

the interrupt issue is related to the rufus-scheduler update after all, the initialization of Scheduler#start
executing a EoTime.now (rufus 3.0 wasn't using a custom Time impl and did a Time.now) which causes local time-zone detection using tzinfo which has some heavy initialization esp. on CI, the scheduler is still stuck in loading .rb files which we try to shut it down.

a work-around of avoiding execution of such code in a separate thread will be in place:

begin
  # TODO: CI work-around - will most likely be moved to the scheduler mixin
  require 'et-orbi.rb' # a dependency of rufus-scheduler since 3.4
  ::EtOrbi::EoTime.now # might take a long time to initialize - loading time zone
  # data (from tz-info) and thus gets un-predictable on CI, since the scheduler worker
  # thread might be stuck starting while we attempt to shutdown in a given time frame
rescue LoadError
end