Webhook calling pipeline trigger on another project renders gitlab instance unresponsive for 10 seconds.
Webhook and trigger should be on the same instance.
Works just like expected when projects hosted on different instances.
Triggered pipeline should contain remote yaml, example:
include:
remote: https://gitlab.instance.example/api/v4/projects/N/repository/files/ci.yaml/raw?ref=REFERENCE&private_token=TOKEN&.yaml
In normal case webhook should get response 201 and pipeline would be started.
In this case webhook gets timeout error but pipeline actually starts.
(You can see timeout error only after gitlab “unfreezes”. “Freeze” time looks exactly the same as time).
Remote can be taken from public repo (no need for token) and problem still occurs.
We reproduces this issue in 2 different on prem kubernetes (official helm) instances and in one omni package in virtual box (Ubuntu 18.04.3) but works just fine in two KVM VMs (omni+ubuntu 18.04.{1,4}) and in Archlinux.
Tested on gitlab 12.2.4 and 12.4.2