Site (an old-style web-chat, no websockets) has one cgi for streaming text chat messages realtime (by javascript code), traffic is very low but work time of this cgi can be hours. There was no problem by now, but last time it stops functioning.
What steps have you taken to resolve the issue?
Disabling “enable-zstd-content-encoding” flag in chrome solved the problem for me, but is it possible to disable Zstd in Cloudflare?
Or maybe tweak it to act as previous compression type, w/o buffering?
Can confirm, this issue started very recently (past days?). This causes website owners to not have any control over output buffering anymore, which in some situations it may be critical.
Manually disabling Zstd in Chrome fixes the issue for self as mentioned by OP, but you can’t control this for others, and the option to disable it in Chrome will probably disappear when they finish the “Zstd experiment”
Hi @pufolog apologies about this. We have identified a bug in ZSTD and Streaming requests. We are releasing a hotfix now and should be complete in the coming hours.
Sorry for the issue caused. I will let you know when this is complete and hopefully should solve the issue.
I’ll be waiting, thank you! I temporary worked it around by replacing streaming to refresh mode, it works, but network traffic and server load increased dramatically.
And a question: if it was done with Server-Sent Events ( Content-Type: text/event-stream), would the problem remains (at current state of this bug)?
We’re still seeing the same issue. We added a rule so we don’t use cloudflare cache for any of our js files (which mostly defeats the cloudflare purpose for us, so hoping this gets resolved soon).
Seems like *.enab.app + enab.app (and also some of another zones). This problem happened yesterday on “ASIACELL ISP” but not with other ISP. The chrome browser shows “The connection was reset.”.
