Language Selection

English French German Italian Portuguese Spanish

New Content/Layout OK?

Varnish Proxy

Silly me, a poll would not work on the new server. I forgot that with the Varnish cache proxy at the front almost all visitors arrive from the same IP address (the proxy), which means that Drupal would allocate just one vote to all (except registered and presently logged in users). With Drupal upgrade we can perhaps find polling software that overcomes this.

rpaf

You must use mod_rpaf to fix this problem that Varnish introduces.
See eg https://www.varnish-cache.org/lists/pipermail/varnish-misc/2008-September/016470.html
mod_rpaf for EL6 64bit here: http://centos.alt.ru/repository/centos/6/x86_64/mod_rpaf-0.6-2.el6.x86_64.rpm

Proxy

Thank, we will look into it. Currently, a lot of stuff other than the poll (e.g. views being counted) are not compatible with Varnish and it makes it look as though not many people visit and can participate in the site.

For sheer stats you could use

For sheer stats you could use an external (i.e. not cached by varnish) service, such as Google Analytics or run your own Piwik.

Piwik

Google Analytics is spyware, but Piwik would be a possibility (Stallman recently told me that it's good). Can it be installed on a cache proxy? I'd have to gain access to it first. Either way, this would not facilitate per-post page request count. Susan had it set up with a module, but it's no longer working correctly. In turn, rating/sorting posts by popularity is no longer possible, and that's the real downside (the front page can no longer list popular items for today).

The problem is not just that IP addresses are not unique. Some requests are never seen by the CMS and Apache.

For the non-unique addresses

For the non-unique addresses look at mod_rpaf, it was made for this situations.
Is this drupal6 or 7? With 6 varnish integration sucks from what I've seen.

See also
https://drupal.org/project/varnish
https://fourkitchens.atlassian.net/wiki/display/TECH/Configure+Varnish+3+for+Drupal+7

Agreed on Google Analytics. You can just install Piwik on the same host and tell Varnish either not to cache it or you can just set its virtualhost on a port other than 80 so it bypasses Varnish completely.

Varnish

Thanks for the pointers.

Yes, it's Drupal 6 and there are other issues that I am beginning to see, such as lack of updates from the RSS feeds around the page (I am currently investigating this, maybe it's related to a cron job or module config although I very much doubt the latter as I haven't changed configs).

Non-unique addresses could be bypassed as an issue even by writing random IP addresses, but that would enable easy poll rigging. I guess it's not essential for operation of the site, but it's a nice-to-have...

From Drupal.org: "This module provides integration between your Drupal site and the Varnish HTTP Accelerator, an advanced and very fast reverse-proxy system. Basically, Varnish handles serving static files and anonymous page-views for your site much faster and at higher volumes than Apache, in the neighborhood of 3000 requests per second."

I have had such issues with Varnish on top of WordPress and MediaWiki (pages served improperly from cache) and it all makes me wonder if removing Varnish altogether is the best way to proceed.

As for Piwik, I have never tried it before, so I will look into it.

I would keep Varnish on for

I would keep Varnish on for static files (css, js, jpeg etc) and to clean up HTTP traffic (Varnish will not forward incomplete or malformed HTTP requests to the backend, it should also be the front line against synfloods etc).

Here's a sample of what I use (test it first, I'm just beginning with Varnish myself)

director default dns {
.list = {
.port = "8080";
.connect_timeout = 5s;
.first_byte_timeout = 600s;
.between_bytes_timeout = 600s;
.max_connections = 10000;
"172.16.1.53"/32;
}
}
sub vcl_recv {
if (req.url ~ "\.(png|gif|jpg|swf|css|js)$") {
return(lookup);
}
}
sub vcl_fetch {
if (req.url ~ "\.(png|gif|jpg|swf|css|js)$") {
unset beresp.http.set-cookie;
}
if (req.restarts == 0) {
if (req.http.x-forwarded-for) {
set req.http.X-Forwarded-For =
req.http.X-Forwarded-For + ", " + client.ip;
} else {
set req.http.X-Forwarded-For = client.ip;
}
}
}

Then install mod_rpaf and make sure your Apache is listening on port 8080 and add this to /etc/httpd/conf.d/rpaf.conf:
LoadModule rpaf_module modules/mod_rpaf-2.0.so

RPAFenable On
RPAFproxy_ips 127.0.0.1 IPs_OF_THE_SERVER
RPAFsethostname On
RPAFheader X-Forwarded-For

PS: looks like drupal is messing with my comments, here's a text version http://fpaste.org/74672/raw/

Thanks

Thanks, I will look at it and into it in the weekend.

RSS feeds

The Piwik demo looks impressive, I have just given them a word of endorsement.

I am still trying to resolve some other issues we've identified.

I think I found the source of the issue above (RSS feeds). It seems like any external site access is denied by default, which helps explain why RSS feeds cannot be retrieved by the Drupal part of the site:


[root@tuxmachines ~]# wget lxer.com
--2014-02-05 04:34:37--  http://lxer.com/
Resolving lxer.com... 108.166.170.174
Connecting to lxer.com|108.166.170.174|:80... failed: Connection refused.
[root@tuxmachines ~]# wget linuxtoday.com
--2014-02-05 04:34:54--  http://linuxtoday.com/
Resolving linuxtoday.com... 70.42.23.121
Connecting to linuxtoday.com|70.42.23.121|:80... failed: Connection refused.

Looks like a firewall issue

Looks like a firewall issue at the first glance.

Firewall

Nux wrote:

Looks like a firewall issue at the first glance.

Yes, it was a simply issue to tackle. It works now.

Pageview count and polls

I'll have a look and see if configuration can solve not just the polling issue but also pageview count. The site of this module is down and it seems like it may require configuration on the cache server too.

More in Tux Machines

Linux Kernel: Chrome OS, Direct Rendering Manger (DRM) and Char/Misc

  • Various Chrome OS Hardware Support Improvements Make It Into Linux 5.3 Mainline

    Various Chrome OS hardware platform support improvements have made it into the Linux 5.3 kernel for those after running other Linux distributions on Chromebooks and the like as well as reducing Google's maintenance burden with traditionally carrying so much material out-of-tree.

  • The Massive DRM Pull Request With AMDGPU Navi Support Sent In For Linux 5.3

    At 479,818 lines of new code and just 36,145 lines of code removed while touching nearly two thousand files, the Direct Rendering Manger (DRM) driver updates for Linux 5.3 are huge. But a big portion of that line count is the addition of AMD Radeon RX 5000 "Navi" support and a good portion of that in turn being auto-generated header files. Navi support is ready for the mainline Linux kernel!

  • Char/Misc Has A Bit Of Changes All Over For Linux 5.3

    The char/misc changes with each succeeding kernel release seem to have less changes to the character device subsystem itself and more just a random collection of changes not fitting in other subsystems / pull requests. With Linux 5.3 comes another smothering of different changes.

today's howtos

Android Leftovers

Spectre Mitigation Performance Impact Benchmarks On AMD Ryzen 3700X / 3900X Against Intel

AMD Zen 2 processors feature hardware-based mitigations for Spectre V2 and Spectre V4 SSBD while remaining immune to the likes of Meltdown and Zombieload. Here are some benchmarks looking at toggling the CPU speculative execution mitigations across various Intel and AMD processors. For this round of testing are some mitigation comparison tests on the Core i7 8700K, Core i9 9900K, Core i9 7960X, Ryzen 7 2700X, Ryzen 9 2950X, Ryzen 9 2990WX, Ryzen 7 3700X, and Ryzen 9 3900X. On each processor, the tests were done when booting the Linux 5.2 kernel with the default/out-of-the-box mitigations for Spectre/Meltdown/Foreshadow/Zombieload (all CPU speculative execution mitigations to date) and then again when making use of the "mitigations=off" kernel parameter for disabling these run-time-toggleable mitigations. Basically the tests are the equivalent of mitigations=off vs. mitigations=auto (default) comparison. Read more