[Ingest Manager] When I unenroll the Agent it doesn't disappear from UI (remains in unenrolling state)

[EricDavisX]

[Ingest Manager] When I unenroll the Agent it doesn't disappear from UI (remains in unenrolling state)

• possibly regressing / coming from this related change: [Ingest Manager] Improve agent unenrollment #18825

For confirmed bugs, please report:

• Version: 7.9 near-BC1 snapshot (from 7/15 afternoon) - kibana commit: e5eadb02
  Agent is from 7/16 Beats CI build.
  image:

• Operating System:
  Win 7 x64

• Steps to Reproduce:
  TL;DR;
  install agent.
  unenroll agent.

...
Details:
I used Cloud to install the stack... I'll post more

api:

before unenroll:

after unenroll:

[elasticmachine]

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

[michalpristas]

do you have agent logs available? did ACK to action arrived?

[EricDavisX]

I hear Nicolas mention he is brewing a PR for this, its reproduced and being handled.

[ph]

PR is available at elastic/kibana#72386

…

On Fri, Jul 17, 2020 at 7:06 PM Eric Davis ***@***.***> wrote: I hear Nicolas mention he is brewing a PR for this, its reproduced and being handled. — You are receiving this because you are on a team that was mentioned. Reply to this email directly, view it on GitHub <#20006 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAAAFAHZTASOJLGN6ISWDPTR4DKPPANCNFSM4O44BPHQ> .

-- -- ph

[EricDavisX]

Hi - I see that the 'force unenroll' is coming in ticket elastic/kibana#72386 , but that's more of a 'save me' option instead of the normal flow of an Agent successfully enrolling, there may yet be a bug here.

@michalpristas here are some logs, from a centos7 unenroll that didnt work:

request looks usual:
Request URL: https://b034192bc7194df8a82b4b440dd1f323.europe-west1.gcp.cloud.es.io:9243/api/ingest_manager/fleet/agents/c0426c1d-59de-40f4-a4a7-f037291aeb96/unenroll

log:
linux-unenroll.log

unenroll...

then stuck for 3 mins, logs were not growing and UI seemed stuck.

I can note that a Win7 x86 host I unenrolled actually worked - if that helps. : /

[ph]

@blakerouse This seems to be a timeout issue that we were discussing this morning.

[blakerouse]

So I looked at this from the Agent side as well, even with timeouts on all 3 sub-applications, endpoint-security, filebeat, and metricbeat my Agent still unenrolled.

I tested this on Windows 10 7.9 BC2.

It did take a bit, probably 2 minutes for it to fully complete once it finished all the timeouts, but it still was marked Unenrolled and was removed from the UI.

[michalpristas]

same here results on my end. i will try few more tests with local agent and cloud UI but i expect same result.

[EricDavisX]

if you're both using BC2 code strictly, then there may be some other factor at play. we could list out what specifics we are doing with the config? the big one is Endpoint. Did you both have Endpoint successfully deployed to the host in your config when you unenrolled? @blakerouse @michalpristas ?

Maybe things will work better in BC3, too - there have been changes to the unenroll and other places in Ingest and a ton on Endpoint and Kibana-endpoint side... we can re-test in a day, if nothing else, and report back

[blakerouse]

@EricDavisX Yes I specifically test with Endpoint enabled and disabled, both worked.