Still having problems logging in and getting the cached "non logged in" version. I just looked at my browser and noticed there is no cookie called "SID" in phpbb3 so you need to make sure what you set is specifically relevant to the actual cookies set by the app. Also you will need to do this so that this only affect phpbb3 and not other apps on the domain. Don't know the exact answer but I am sure with some google time and/or posting in the nginx mailing list, you'll be able to figure it out.

PS. you might need to change the proxy_in_redirect setting back to the default. I can't recollect which one is supposed to remove the port - check docs.


The fallback 404 stuff is not compulsory. It was useful for a specific application I was using where I knew that just before Nginx could not find something does not mean it doesn't exist (a so called "image firewall" in Gallery 2) so when it returned "404" I passed it to Apache.


Fanboy? Hmmm. Anyway. People will always have to make modifications on their own and I think the best you can do is give a generic config similar to the one I posted a while back and user will have to manually modify as required as it is simply not possible to provide a one size fits all config.

I seem to recollect sage words of wisdom from you to this effect when I asked whether the ART Nginx install package could set up so that it starts on reboot. You said then that it had to be done so that users have to set this up themselves. I would have thought that making it start on reboot is infinitely more straightforward than a global, every possibility config.

On the test environment, I am setting up a linode to try out Cherokee and also a switch to Ubuntu instead of Centos but this is tied up for the moment. Perhaps you could give the mailing list a shot.


License:- Creative Commons
mod_sec:- No. (but don't tell Mike )
ioncube:- It should since this is a php thing
zend:- Yes
etc:- Not sure

I haven't tried it yet but after playing with Nginx for a few months, I am looking to something else to become a fan of.

1) Ok I know why this doesnt effect me now. I'm using SSL so I'm bypassing nginx. We need a more sticky solution, one off config settings on a per app basis wont scale to a bigger hosting environment. Even if the end admins have the skill to do it, they may not have the time.

2) Ok so the 404 stuff is (more or less) off the list at least. Hooray, 1 down. When I tested this on a site that had custom 404s they worked successfully.

3) Ha! Thats what they said about virtual hosting in 1997. I think we can solve this.

In this case setting nginx to start at boot without atomic-accelerator (AA) was the issue (because it would start on port 80, which already had plesk). The solution was to add the second "meta" package (AA) to handle all the configs to everything else (apache, mod_rpaf, plesk). So sticking all that glue in the AA package is an option now, we can do pretty much anything we want. Right now AA will install nginx, move apache to 8080, configure mod_rpaf with the IP of the server, and insert the events into plesk. Even if this needs to be something like a transparent firewall rule (a la squid) thats on the table. If we have to put a front end on nginx to make it easy for a user to customize it, we can do that too.

4) Once we get the nginx stuff licked we can give cherokee a shot.

That explains it. With regards scaling, the more finely tuned Nginx config are, the less they can be used as one size for all. While it can be used for such, Nginx is not geared for mass hosting in the first place.


Yeah. Just an example of a specific config for specific uses.


Which is why I called your words sage. I know the yum install *could* result in a collision on Port 80.

On front ends, I got Cherokee going and it has a user easy web based admin interface ... very 2010, you'll love it. Leaves me a bit uneasy though as I believe such stuff should done in text files using nano all the time.

You mean vim. Nano is the devil.

Is NGINX / Atomic Accelerator recreating /etc/httpd/conf.d/zzz_atmail_vhost.conf periodically?

I removed this file, it came back. I cp /dev/null > zzz_atmail_vhost.conf / chmod 000 zzz_atmail_vhost.conf -- and today it's back.

I have atmail 'installed', but swapped Roundcube in its place and am having my roundcube hack disabled daily.

It would trigger events if it was enabled/disabled through plesk. Updating it via plesk might do it too.

that may be part of the plesk daily cron job, I think all of those zzz files are Plesk maintained, so you cant edit those as the changes will probably get wiped out.