I’ve run into an issue where several domains/sites return 0 values for key metrics like First Byte, with very high load times (5+ seconds).
23
Lighthouse PWA Score
N/A
First Byte Time
N/A
Keep-alive Enabled
N/A
Compress Transfer
N/A
Compress Images
N/A
Cache static content
X
Effective use of CD
Taking a look at the only waterfall request metric available suggests it’s due to web sockets traffic over wss:// protocol. Screenshots are attached.
I run a test for the same exact domain/site here on webpagetest.org, and don’t see the same sockets requests recorded in the waterfall. The tests also complete normally.
Switching to Firefox or some other browser for the test on our private instance also seems to allow the full test to complete. That seems to isolate the problem to just Chrome agents.
My assumption is that this public instance prevents or deemphasizes sockets traffic.
- Is that assumption correct?
- How can I configure our private instance to behave the same?
We need to allow testing over Chrome agents, so I’ve been blacklisting certain services, like nexus-websocket-a.intercom.io, for the time being. That doesn’t seem to be a scalable solution in the long run, however, so looking for config alternatives.