"no data received" with php-fhm

[jeffkaufman]

khmer365 wrote:

When ngx pagespeed loaded, the website gave "No data received.". I can't
figure out what's wrong.
...
/etc/nginx/nginx.conf -> https://gist.github.com/4617546
/etc/nginx/conf.d/server.conf -> https://gist.github.com/4617556
/etc/nginx/php -> https://gist.github.com/4617563

I use nginx with php-fpm. When I turn pagespeed off, everything went back
fine. There is nothing in the log.

Replicate the setup and see whether we work properly with php-fhm.

[jeffkaufman]

Testing with php-fhm on my machine we seem to work fine. Asked for clarification on the mailing list.

[khmer]

I turned warn to debug but still nothing in the error log.

[jeffkaufman]

On the mailing list I suggested testing with pagespeed off in the config, which worked. This disables all rewriting at a very low level, so the next thing to test is pagespeed on with pagespeed RewriteLevel PassThrough, removing any pagespeed EnableFilters directives. I suggested this on the mailinglist, still waiting for a response.

[oleander]

I'm having the same problem.

Tried using pagespeed RewriteLevel PassThrough but I'm stil getting random "No data received" errors. I'm also seeing a lot of NgxBaseFetch::RequestCollection: Broken pipe errors in /var/log/nginx/error.log.

Any solution to the problem? I'm running X-Page-Speed: 1.5.27.3-3005 with nginx version 1.4.1.

[jeffkaufman]

When you say "random" do you mean that the same url sometimes works with passthrough and sometimes gives "no data received"?

[oleander]

@jeffkaufman Yes, that's correct. A note tho, I'm not using php-fhm.

[jeffkaufman]

That's strange. Do you see anything in your error log when this happens? Anything unusual about the requests that fail?

Is the failure rate 90%? 10%?

[oleander]

I don't see anything at all in the error logs and I can't reproduce the error. I've seen the error 3 times (out of ~ 200 requests). The error NgxBaseFetch::RequestCollection: Broken pipe occurs once every other request tho.

[oleander]

I just saw the error again, without pagespeed being *active. The error log is still empty. Just reinstalled nginx 1.4.1 without pagespeed.

• pagespeed on; in nginx.conf was removed from the file, which I guess is the same thing has pagespeed off;

[jeffkaufman]

Strange. ngx_pagespeed does almost nothing if it's disabled, not installing any handlers. With ngx_pagespeed uninstalled are you no longer seeing those occasional "no data received" messages?

[oleander]

I'll get back to you on that one.

25 jun 2013 kl. 17:44 skrev Jeff Kaufman notifications@github.com:

Strange. ngx_pagespeed does almost nothing if it's disabled, not installing any handlers. With ngx_pagespeed uninstalled are you no longer seeing those occasional "no data received" messages?

— Reply to this email directly or view it on GitHub.

[oleander]

Just saw the error again (with pagespeed not even being compiled into nginx), so it's not related to pagespeed. I'll try to down grade to nginx 1.3.x.

I'll keep posting.

[jeffkaufman]

Stress testing your php install might help debug this. It sounds like every so often it fails to respond.

[oleander]

I'm using Ruby (unicorn) under the hood, but I guess the same principas are applied. I just downgraded to 1.2.6. I'll try to run it a production for a week and then activate pagespeed again.

I'll post the results as soon as I've some.