Like Ra's Naughty Forum

Full Version: Site performance, Server errors, outages and tunings
You're currently viewing a stripped down version of our content. View the full version with proper formatting.
Nice... One of the reasons I prefer CLI over GUI. Last night I added a AAAA DNS entry. And no, I did not forget the trailing dot for the name. But the GUI ignore the dot. Ha-ha ... Also it turned out, that the DNS manager does not accept @ as the domain name (but it uses @ in the zone file), but expects empty name to refer to the domain.

It costed me hours of downtime and cache poisoning.


Crap.... Why does not it just work? 😁
The new zone is still not propagated everywhere. E.g. Google, gtmetrix cannot yet connect.
The forum has been upgraded to the latest version. One plugin is incompatible (View unread post).

Upgrading plugins...
Lots were changed in 1.6.5. Let me know if something is broken.

Removed the AAAA record from the DNS. I wonder if the glitches with resolving are related to IPv6.
It was, indeed, the AAAA DNS record. After I removed it Android can see the site again.
The site design is broken in the latest Firefox Beta Mobile, no problem in the native Android browser.
Dunno what happened. Just like earlier: at 14:00 GMT all the memory was used and swapping killed the VPS. Rebooting helped.

How do I limit the amount or running apache processes? Both HTTP daemons and PHP? Apparently whatever I do does not work.
Hmm,

how did you configure your apache2 and php?

You should use the mpm-woker module and with that fast-cgi (fcgi). Use google to find some infos on that.

This http://www.marco-rust.de/worker-vs-prefork-mit-php/50/ is a cool site describing how to tune that combo.
It is in German, but if you have some trouble with understanding ask me - I ll help you

bye
krinlyc
(18 Dec 2011, 10:09 )krinlyc Wrote: [ -> ]You should use the mpm-woker module and with that fast-cgi (fcgi).
Not using fast-cgi. But you are right about mpm. This is exactly what was wrongly configured. Bobbi (https://the.tiesthatbind.me/ ) and I figured that out yesterday. all settings were inside the MPM tags, but no MPM module was installed. So, I moved them outside of the tags, decreased the MaxClients even more ( it's 50 now) and here's the result:

[Sun Dec 18 13:25:58 2011] [error] server reached MaxClients setting, consider raising the MaxClients setting

It works! And the OS was not out of memory as before.
(18 Dec 2011, 15:01 )Like Ra Wrote: [ -> ]
(18 Dec 2011, 10:09 )krinlyc Wrote: [ -> ]You should use the mpm-woker module and with that fast-cgi (fcgi).
Not using fast-cgi. But you are right about mpm. This is exactly what was wrongly configured. Bobbi (https://the.tiesthatbind.me/ ) and I figured that out yesterday. all settings were inside the MPM tags, but no MPM module was installed. So, I moved them outside of the tags, decreased the MaxClients even more ( it's 50 now) and here's the result:

[Sun Dec 18 13:25:58 2011] [error] server reached MaxClients setting, consider raising the MaxClients setting

It works! And the OS was not out of memory as before.

If you use the new mpm module you should use fcgi. Otherwise it is somehow as if you drive a Porsche in second gear....

And even more important is the fact that mpm is not threadsafe; mod_php is not aware of this fact.

krinlyc