[04:08] <foo> Er, someone is telling me how debian 2.6.8 is better than ubuntu 2.6.16.
[04:08] <foo> He is telling me to try a 2.6.8 kernel on ubuntu
[04:16] <infinity> I wouldn't recommend it.
[04:16] <foo> eh, yeah, we're not going to do that.
[04:16] <infinity> OTOH, we've never had a stable release with 2.6.16 either.
[04:16] <foo> I know there will be some issues with udev, I think.
[04:17] <infinity> breezy = 2.6.12, dapper = 2.6.15, edgy = 2.6.17, feisty = 2.6.20 (when it's out)
[04:18] <foo> eh, I'm just having some issues. Everyone says it's IO, but the data center doesn't accept that answer.
[04:18] <foo> This server pushes 100mbit .. and the iowait from iostat increases.. 
[04:18] <foo> Load ranges from .1 to 350
[04:29] <infinity> What does it serve?
[04:29] <foo> files with apache, mainly.
[04:29] <foo> Well, only.
[04:29] <infinity> Static, or dynamic?
[04:29] <foo> dynamic
[04:29] <foo> lots of file uploading/downloading
[04:29] <foo> 100mbit cap
[04:29] <infinity> DB backend?
[04:29] <infinity> PHP, Perl, Python?
[04:30] <infinity> It's entirely possible you just allow too many client connections.
[04:30] <infinity> Alternately (and this one sounds weird until you think about it), you may be running too few apache instances, not too many.
[04:30] <infinity> It's very costly for apache to fork new children when the hit count increases.  Very.
[04:30] <infinity> If you start with more MinSpares, it goes more smoothly.
[04:31] <foo> PHP
[04:31] <foo> Hm. infinity, should I check for connections about netstat or something?
[04:31] <foo> I've had several guys look at this ... everyone is blaming IO
[04:31] <infinity> When the load is huge, "ps ax | grep apache" and see how many are running.
[04:31] <foo> But how could a network cap at 100mbit cause IO on a server.. I'd expect it to read/write with that
[04:32] <infinity> If there's several hundred, or something, then you really want to bump up MinSpares in the config, or limit MaxClients.  Pick one.  Either will help.
[04:32] <foo> Apache has already been optimized, but. hm
[04:32] <infinity> Err, when people blame "I/O", they're not blaming the network, they're blaming slow disk and such.
[04:33] <foo> # netstat -nap|grep -c ESTA
[04:33] <foo> 381
[04:33] <foo> yeah, I know
[04:33] <foo> I just don't get how a server pushing 100mbit could have IO
[04:33] <foo> issues
[04:33] <infinity> Why not?
[04:33] <infinity> It's not about raw disk throughput, it's about crazy random access.
[04:33] <foo> heh, that might be what it is
[04:33] <foo> That's what I'm thinking
[04:34] <infinity> But in my experience with PHP/Apache, I'd suspect you're CPU bound on apache forking.
[04:34] <foo> 192 requests currently being processed, 20 idle workers
[04:34] <foo> load is 1.25 right now
[04:34] <foo> Pushing 81mbit right now
[04:35] <infinity> grep SpareServers /etc/apache2/apache2.conf
[04:35] <foo> MinSpareServers      20
[04:35] <foo> MaxSpareServers     27
[04:35] <foo> worker and prefork, probably
[04:35] <infinity> No, those are both for prefork.
[04:35] <foo> oh, ok
[04:36] <infinity> Min is the number started at startup as spares, Max is the number it will keep idle once connections die off.
[04:36] <foo> ok
[04:36] <infinity> For your sort of load, you might want something more insane like Min 50, Max 75
[04:36] <infinity> Or even higher.
[04:36] <infinity> If you're doing 200 requests right now.
[04:36] <infinity> Min 100, Max 150.
[04:36] <infinity> Experiment, have fun. :)
[04:37] <infinity> If you have the RAM to back up that sort of thing anyway, which you better if you're trying to saturate 100Mbit.
[04:38] <foo> Hm, 1GB RAM
[04:39] <foo> infinity: FYI, I've been playing around for weeks now. I've pretty much had it. 
[04:40] <infinity> Is that PHP serving content from a DB?
[04:40] <infinity> If so, tuning said DB enging for that sort of load is just as important as tuning Apache.
[04:41] <infinity> s/enging/engine/
[04:42] <foo> Nope, not from DB
[04:42] <infinity> Kay.
[04:59] <ArwynH> lo
[05:00] <ArwynH> just wondering, is there any plans to implement a server control panel of sorts?
[05:01] <ArwynH> If so, can I have a link to the spec please? Or should I write one myself?
[05:54] <levander> ArwynH: there are already some of those written that are free software, like cpanel
[05:55] <foo> ArwynH: Look into webmin
[06:17] <ArwynH> foo: webmin is a nightmare and is not in repos.
[06:17] <foo> ArwynH: Nightmare? Why? Yeah, isn't in repos.
[06:18] <ArwynH> foo: it isn't in the repos because the code is a nightmare to maintain. I've tried running it before.
[06:19] <ArwynH> it's a pain to run aswell
[06:19] <ArwynH> anyway, a web interface isn't what I'm looking for.
[06:19] <foo> I see
[06:21] <ArwynH> I've pretty much decided what I want and I havn't found anything like it in the oss world, so It looks like I'll have to write it myself, but I'm just trying to make sure.
[06:21] <foo> ah
[06:25] <levander> ArwynH: maybe write a spec first and circulate, maybe people will help you based on the spec?
[06:25] <ArwynH> levander: i plan to. hence me checking if there was one available to look at first.
[06:26] <levander> ArwynH: maybe circulate it on webmin and cpanel mailing lists, after written?
[09:29] <okaratas> hi
[10:59] <shwag> phpmyadmin is pretty out of date on ubuntu.
[10:59] <shwag> 2.8.0.3  vs   2.9.1.1 
[11:54] <linuxpoet> why is mailman on dapper always using the wrong domain to send?