WhatsNew Controller looking for data in the incorrect location on Kubernetes installations
Summary
It would appear that data being served out of our CNG containers is looking at the path associated with omnibus instead of the correct location for where GitLab is installed in the container images. This is leading to 2 problems:
- We are seeing an HTTP500 - this is odd considering it is serving a file that it can't find, I would expect this to be a 404
- We are not serving the data despite it existing and is readily available on the Pod.
git@gitlab-cny-webservice-web-67f4d5d4fd-zxbd4:/srv/gitlab/data/whats_new$ ls
202008180001_12_10.yml 202009150001_13_03.yml 202012160001_13_07.yml 202104220001_13_11.yml templates
202008180002_13_0.yml 202009300001_13_04.yml 202101140001_13_08.yml 202105220001_13_12.yml
202008180003_13_01.yml 202010230001_13_05.yml 202102180001_13_09.yml 202106220001_14_0.yml
202008210001_13_02.yml 202011230001_13_06.yml 202103220001_13_10.yml 202107220001_14_1.yml
For some reason, the container is attempting to look here: /opt/gitlab/embedded/service/gitlab-rails/data/whats_new/...
Steps to reproduce
Bring online a helm install GitLab, navigate to /-/whats_new
and observe an HTTP500 on installs that are served by Kubernetes.
What is the current bug behavior?
An HTTP500 while serving data located here: https://gitlab.com/gitlab-org/gitlab/-/blob/master/data/whats_new/202105220001_13_12.yml
What is the expected correct behavior?
An HTTP200 while serving any data located here: https://gitlab.com/gitlab-org/gitlab/-/blob/master/data/whats_new/
Relevant logs and/or screenshots
- https://sentry.gitlab.net/gitlab/gitlabcom/issues/2756907/?query=is%3Aunresolved%20correlation_id%3A01FCVE7Y9KZ9Y4BBQV9KX49DA1
- https://log.gprd.gitlab.net/goto/6d417cca1bdd90ab70a314326077d32e
Output of checks
This bug happens on GitLab.com