I've tested ngx_pagespeed on multiples configurations and found that on about 50% of instances (randomly) encountered a problem with firefox on www.pratique.fr :
I've never been able to reproduce the problem with Chrome.
It seems that time-outs happen when loading the page.
The text was updated successfully, but these errors were encountered:
I got this to reproduce on chrome with the developer-tools networking tab open. It shows the html download as failed:
Looking at the response body, it appears to have only downloaded part of the html, cutting off after:
Trying again it cuts off in a different place:
Is there anything unusual in your server logs?
You html is on the large side (200k) and it's full of large base16 classes (ex 45CBCBC02D1F1FCCCCCC19C0C14ACB46CAC3431944C11F484F424ACB464F491E4E43C3414843431E424A4BC1431E4DC3C1464B46CAC3431945CB4E48) but those shouldn't break ngx_pagespeed.
Aha! I can reproduce this with curl as well, but only if I give
I'm seeing failures only some of the time.
Downloading the html and testing on jefftk.com I see the same problem:
Here's the error log at debug:
I think there aren't any other requests intermingled, but I'm not completely sure.
Strangely enough, I also see this problem when trying to download the error log itself over http:
Bizarrely, this time it does work with gzip:
The error log from when download the error log fails is:
I am extremely confused.
I just got an idea: when we reordered our module to run in front of the gzip module, we ran in to trouble when passing on NULL cl's, remember? Maybe that code is worth having a look at regarding this problem. I can't do that right now, but I should be able to do so this evening.
On our logs we have that kind of errors:
NgxBaseFetch::RequestCollection: Broken pipe
2013/05/04 10:25:48 [alert] 9296#0: [ngx_pagespeed 22.214.171.124-2912] [0504/102548:FATAL:proxy_fetch.cc(79)] Check failed: outstanding_proxy_fetches_.empty().