Persistent 502 error just in the SIdeKiq panel since upgrading to gitlab-ce 8.14.x

Summary

We experience a Gitlab SideKiq 502 error page in the inner "frame" where the Sidekiq panel is supposed to appear in the Admin area of gitlab.

sidekiq_502

Steps to reproduce

It seems some local configuration issue may be required to reproduce it.

Expected behavior

Able to open sidekiq.

Actual behavior

Unable to use sidekiq.

Relevant logs and/or screenshots

  • Nothing appears in /var/log/gitlab/gitlab-rails/sidekiq.log when I load the sidekiq admin page. I watch it with tail -f.

  • Using gitlab-ctl tail I was able to observe this oddity:

==> /var/log/gitlab/nginx/gitlab_error.log <==
2016/12/06 12:29:35 [error] 1340#0: *36 upstream sent too big header while reading response header from upstream, client: 192.168.215.154, server: gitlab.ramsoft.biz, request: "GET /admin/sidekiq HTTP/2.0", upstream: "http://unix:/var/opt/gitlab/gitlab-workhorse/socket:/admin/sidekiq", host: "gitlab.ramsoft.biz", referrer: "https://gitlab.ramsoft.biz/admin/background_jobs"

  • It occurs to me to wonder why gitlab_error.log isn't one of the longs shown in the Admin logs section of the UI?

  • Google fu leads me down the NGINX requires magic config settings to Just Work.