[Fleet / Agent] when persistent agent 'install' is rebooted on darwin system, not all Beats are restarted

[EricDavisX]

[Fleet / Agent] when persistent agent 'install' is rebooted on darwin system, not all Beats are restarted

related to #21424 which is for the Elastic Endpoint port. This is just for the other beats, tho the cause may be related (tbd), logging separately for safety.

tested with latest 8.0 Agent + Kibana
agent:
https://snapshots.elastic.co/8.0.0-f437c552/downloads/beats/elastic-agent/elastic-agent-8.0.0-SNAPSHOT-darwin-x86_64.tar.gz

kibana, commit: 5aa3fb5e0634c84ae3521de6cad663585d7b01ae, from date: Wed Sep 30 19:04:22 2020 -0600

Tested on darwin 10.15.6 (19G2021)

download the agent and unzip.
install kibana via cloud
use the (not yet shown in the UI) steps to install, as:

then add Endpoint to the policy
then wait 5-10 mins for it to all come on line
do a 'ps ax | grep elastic' to see all of the Beats running, from the default config options (4)
then reboot the machine.
see that only 1 beat comes back with Agent (Endpoint concern is logged separately)

screenshots:
before reboot, after install and policy change and 5 min wait:

after reboot:

note we have a test that covers this, but not sufficiently - we need help to implement the test, see:
elastic/e2e-testing#218

[elasticmachine]

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

[ph]

@EricDavisX This is with a latest snapshot?

@blakerouse Can you take a look?

[EricDavisX]

Yes, PH.

Here are the logs, and I acknowledge some errors there are known, and even expected since I ran against my own laptop (unadvised) and Mac SIP protections are in place (which explains the failed Endpoint policy)

darwin-agent.log

[EricDavisX]

@rahulgupta-qasource please hold off on doing more install tests for reboot until after this is fixed. you can still test basic initial stand-alone mode install, and we can test 'upgrade' when its available (soon).

[ghost]

Hi Eric

Thank you for the update.

We have hold on the install tests for reboot validated the basic initial stand-alone mode install under #21019 ticket.

We will test the 'upgrade' tests once they are available.

[blakerouse]

@EricDavisX Is this still reproducable? I think this was because of the action_store.yaml issue that is now fixed.

[EricDavisX]

indeed this looks fixed to me - tested on 7.10 BC2 on a catalina darwin host. i feel ok to close this out.