GitLab Pages - Outage 10:15 PM UTC 2017/03/10
Overview
-
GitLab pages
HTTPS
redirects fail in an infinite loop - 10:15 PM UTC 2017/03/10 -
@yorickpeterse mentions this on Slack https://gitlab.slack.com/archives/infrastructure/p1489184308193924
-
@stanhu mentions a fix for the configuration
/opt/gitlab/sv/gitlab-pages/run
- https://gitlab.slack.com/archives/infrastructure/p1489185831468207 - - 10:45 PM UTC 2017/03/10# diff run.orig run 27,28c27,28 < -redirect-http=false \ < -use-http2=true \ --- > -redirect-http false \ > -use-http2 true \
-
This fix is deployed but the problem persists.
-
Stan mentions an issue with
app.go
if https && !a.RedirectHTTP
- https://gitlab.com/gitlab-org/omnibus-gitlab/merge_requests/1348#note_25156832 - 10:53 PM UTC 2017/03/10 -
This is rolled into a new diff patch https://gitlab.com/snippets/1650007/raw - 1:35 AM UTC 2017/03/10
-
Patch deployed - 1:45 AM UTC 2017/03/10
-
Problem persists. Investigation reveals gitlab-pages is refusing to restart with the standard
gitlab-ctl restart command
- 2:00 AM UTC 2017/03/10 -
GitLab pages is force restarted
sudo kill -9 $(ps aux | grep '[g]itlab-pages-unprivileged' | awk '{print $2}') && sudo gitlab-ctl start gitlab-pages" -t 5
- 2:20 AM UTC 2017/03/10 - Problem resolved.
Problems
-
Incorrect conditional statement added 1 year ago https://gitlab.com/gitlab-org/gitlab-pages/commit/3d691bf42432304a3ae6d6465c3f596629891e74
-
Not sure how this changed on production as the config for
cluster
and the newweb
works was http redirect = false