Hi community,
We’ve encountered an intermittent issue with WebSocket connections when routing traffic through Fastly. The behavior is consistent and reproducible mostly in Google Chrome, while all other browsers (e.g., Safari, Firefox, Brave) work as expected.
Issue Summary
When a WebSocket connection is first established, it successfully upgrades (HTTP 101) and works normally. However, any subsequent WebSocket connection attempts from the same client fail to open.
This pattern repeats consistently:
-
The first connection always succeeds.
-
Subsequent attempts fail immediately.
-
The issue occurs mostly when requests are routed through Fastly.
-
When bypassing Fastly and connecting directly to the origin, all WebSocket connections work properly across all browsers.
Technical Details
-
In the failing cases, the connection attempt completes in about 60–70 ms, which is far too quick for a genuine WebSocket handshake.
-
It feels like these connections are being interrupted or terminated early, somewhere within the Fastly layer or a related intermediary component, rather than being passed through to the origin server.
-
The difference in timing is significant: failed connections complete in ~60–70 ms, while successful handshakes take ~1.6–1.9 seconds. This indicates that the failed requests likely do not complete the full upgrade sequence.
Request
Could you please let us know if this is a known issue or if there have been any recent reports of WebSocket connection failures or timing anomalies when routed through Fastly?
Any insight into potential root causes or conditions that could trigger this behavior would be greatly appreciated.
Thank you for your help and time!