On gitlab.com GitLab Pages projects, 502 errors are intermittently returned

I tried to address this via Twitter at https://twitter.com/jpluimers/status/1479192845588307975

Hi @gitlab, I configured pages for a http://gitlab.com public repository with a custom domain which mostly works fine, but every couple of days it will return a 502 error on loading. Refreshing it later then will reload correctly.

What should I investigate further?

Since no response there, here is a case where it goes wrong intermittently every few days (but not right on the clock).

It is the web-dashboard for my mentally retarded brother. It refreshes every 3 minutes as the underlying widgets cannot refresh automatically, or sometimes fail to refresh.

Every couple of days (but no exact frequency), the page returns a 502 error.

This happens on multiple internet connections (all from The Netherlands):

  • xs4all fiber from his home
  • Helden van Nu fiber form my home
  • xs4all fiber from my home
  • Ziggo cable from a friends home

httping statistics:

httping --count 10 https://martijn.pluimers.com
Auto enabling SSL due to https-URL
PING martijn.pluimers.com:443 (/):
connected to 35.185.44.232:443 (363 bytes), seq=0 time=383.29 ms 
connected to 35.185.44.232:443 (363 bytes), seq=1 time=385.38 ms 
connected to 35.185.44.232:443 (363 bytes), seq=2 time=417.22 ms 
connected to 35.185.44.232:443 (363 bytes), seq=3 time=460.04 ms 
connected to 35.185.44.232:443 (363 bytes), seq=4 time=505.75 ms 
connected to 35.185.44.232:443 (363 bytes), seq=5 time=373.62 ms 
connected to 35.185.44.232:443 (363 bytes), seq=6 time=303.18 ms 
connected to 35.185.44.232:443 (363 bytes), seq=7 time=300.45 ms 
connected to 35.185.44.232:443 (363 bytes), seq=8 time=401.31 ms 
connected to 35.185.44.232:443 (363 bytes), seq=9 time=298.93 ms 
--- https://martijn.pluimers.com/ ping statistics ---
10 connects, 10 ok, 0.00% failed, time 13866ms
round-trip min/avg/max = 298.9/382.9/505.8 ms

Since it is hosted on gitlab.com, I cannot check the status of https://docs.gitlab.com/ee/administration/pages/#502-error-when-connecting-to-gitlab-pages-proxy-when-server-does-not-listen-over-ipv6 nor https://docs.gitlab.com/ee/administration/pages/#intermittent-502-errors-or-after-a-few-days (but I presume those are configured correctly, right?)

For his personal dashboard machine, there is a cumbersome workaround (he mentally cannot tell proactively when an error page shows as usually it bumps him into a "red" status, see https://blog.chaddickerson.com/2019/01/30/the-magic-of-the-personal-check-in-red-yellow-green/): regularly start a VPN connection to his home, then via VNC check the dashboard state, and manually refresh the dashboard when the 502 error has occurred.

There is a high chance my metastases will return (see my Twitter profile), so before things get worse I want to leave a situation behind that requires as little manual action as possible.

What can I do to help investigate this further?

Note: I have just started to check https://wiert.me.gitlab.io/private/web-sites/martijn.pluimers.com-ssl/agenda-month-weather.html from a VM to see if the error occurs there as well. For now I resume it does.

--jeroen

Edited by Jeroen Wiert Pluimers