Restful endpoints getting ERR_HTTP2_PROTOCOL_ERROR

Earlier today almost exactly when the Cloudflare Dashboard went down, we started getting warnings that our site was unreachable. It appears that some of our restful API endpoints are suddenly failing. In Chrome we’re getting an ERR_HTTP2_PROTOCOL_ERROR and doing a --trace with CURL shows us that the response is suddenly being cut off with a “HTTP/2 stream 1 was reset” error. Strangely enough, only our restful endpoints seem to be affected, not static assets like images or HTML.

Disabling the Cloudflare DNS proxy and connecting directly to the server fixes the problem, but is of course not an actual reasonable solution for a number of reasons. So far, everything we can find points towards it being a Cloudflare issue.

The best way to look at this more is with a netlog, could you capture one showing the issue? It will help show where and why, I’d also recommend trying with http3 off, http2 off, http1 off and trying to isolate it to which HTTP version exactly this is happening for (you can still see a HTTP2_PROTOCOL_ERROR in h3 and so on)

(Also just to be clear, it will not be related to the dashboard issue. Proxying traffic and the dashboard are very different systems and do not relate)

1 Like

Alright, it appears to be a false alarm. It appears that one of our http headers had been mangled which somehow made the Cloudflare proxy fail. Why it happened to coincide with the technical issues Cloudflare had today is unclear, but might just have been bad luck or maybe some cache that happened to be refreshed at the same time.

Anyway, I believe the issue has been solved.

1 Like

This topic was automatically closed 2 days after the last reply. New replies are no longer allowed.