@dmakovey example pipeline: https://dev.gitlab.org/gitlab/omnibus-gitlab/-/pipelines/425571
The manifest-upload job runs a few stages prior to the check-packages-availability trigger job. This appears to become a blocking job,
Could we add an empty needs that way there's no dependency on the pipeline stage?
Also question. I'm noticing this rule:
- if: '$PIPELINE_TYPE =~ /_(RC|TAG)_BUILD_PIPELINE$/'
when: manual
Is on this job. What plays those? I'm not aware of release-tools performing this action
John Skarbek (97fef25f) at 17 Mar 14:03
John Skarbek (cd06dda9) at 17 Mar 14:03
Merge branch 'sk/update-mirrors-doc-release-platform' into 'master'
... and 1 more commit
Updates the security mirrors documentation to recommend the new YAML-driven Release Platform module as the preferred method for creating security mirrors.
<details> blockThe Release Platform module (work item #50) provides a standardized, YAML-driven way to provision the complete 3-mirror architecture with built-in integrations for common-ci-tasks, Vault, and more. This simplifies onboarding new projects and ensures consistent configuration.
Ok, after a bit of research, I see no evidence that this is actually true
@myacksmith I'm simply copying and pasting the work that you've done. Did you attempt apache style annotation at all? I'm willing to give this a shot. Curious your take prior to moving forward and accepting such suggestion.
John Skarbek (4a3f776d) at 16 Mar 20:03
John Skarbek (0a52bcf3) at 16 Mar 20:03
Merge branch 'sh-gprd-cny-load-shedding' into 'master'
... and 1 more commit
This enables the Workhorse load shedding feature added in gitlab-org/gitlab!218865. In gprd, we use the following Prometheus query to determine that a backlog of 50-100 might be a reasonable threshold to start:
max by (pod) (
max_over_time(puma_queued_connections{type="api"}[5m:15s])
)
In addition this enables NGINX to retry on 504 errors up to 3 times.
Relates to gitlab-com/gl-infra/production#21451
Please read the Contributing document and once you do, complete the following:
This is ready for review!
Configure gunicorn_access_logformat for both API and Content pods in the pre environment to output structured JSON logs for better observability and log parsing.
remote_ip, method, path, status, duration_us, correlation_id
remote_ip, request, status, duration_us, correlation_id
Worker pods run Pulp's distributed task worker process (not gunicorn), so gunicorn_access_logformat is not applicable to them. Workers process asynchronous tasks from a queue, not HTTP requests.
Relates to https://gitlab.com/gitlab-org/build/team-tasks/-/work_items/63