Badgen.net: Bad Gateway errors

[martinklepsch]

Hello & thanks for this fantastic service!

I've recently been getting more and more 502 Bad Gateway errors from badgen.net. I use badgen for the badges of cljdoc.org where I provide a static URL that people can put into their READMEs, e.g. https://cljdoc.org/badge/nrepl/nrepl.

The error page links to this https://vercel.com/docs/error/application/NO_RESPONSE_FROM_FUNCTION

To do this I'm requesting the badge from a server and then respond to the request with the body that badgen provided and some extra caching directives, the code can be found here (Clojure)

My question is:

• Do you have any idea why I might be seeing those 502 errors?
• Should I be calling the badgen API in any different way?

[amio]

Usually it could be error response for live badge, but not for static badge. I've seen some 502 error with newly created badge, but after a refresh it works fine (looks like a cache issue).

Could you provide some error url?

I got some 502 logs, but mostly during the david-dm outage recently.

[martinklepsch]

I made some changes to get better reporting and will report back. All the URLs include the string "cljdoc" if that's helpful.

[wopian]

Was going to open an issue to gracefully handle 5xx responses from API providers (in this case David DM).

When these APIs return 5xx, the badge endpoint returns a HTML error page instead of a SVG which results in the following with this (https://flat.badgen.net/david/dev/wopian/kitsu) badge endpoint.

Ideally it would show something akin to

[amio]

Yep, we should handle 500 gracefully

[wopian]

The GitHub Checks badge for only one specific repo of mine is also returning a badgen: error badge with an intermittent HTML 502 response while the latest commit has unfinished checks

https://flat.badgen.net/github/checks/wopian/kitsu

[amio]

"unfinished checks" 🤔

Looks like this is the root case we are looking for 👍

[phoenixeliot]

I'm getting a lot of these looking through the examples on the site, as well. Maybe 20-30% of the examples I opened had this problem, but only on the first request time I opened each one (I'm assuming there's some kind of caching going on?)

[amio]

I'm getting a lot of these looking through the examples on the site, as well. Maybe 20-30% of the examples I opened had this problem, but only on the first request time I opened each one (I'm assuming there's some kind of caching going on?)

I also noticed this issue, but I cannot reproduce it 🤔
Guessing it's caused by a cold start without cache (after new deployment)?

[amio]

Did a test after new deployment, still no issue