--- Day changed --- Log opened Mon Apr 24 23:59:03 2006 00:08 -!- OverlordQ [erwin@ip70-178-103-82.ks.ks.cox.net] has joined #linode-xenbeta 00:09 -!- OvrLrd-Q [erwin@ip70-178-103-82.ks.ks.cox.net] has quit [Read error: Connection reset by peer] 00:10 -!- SupaZubo1 [~crack@frotz.zork.net] has joined #linode-xenbeta 00:10 -!- SupaZubon [~crack@frotz.zork.net] has quit [Read error: Connection reset by peer] 00:10 -!- tierra [~tierra@ibaku.net] has joined #linode-xenbeta 00:28 -!- SupaZubo1 is now known as SupaZubon 01:17 -!- Marcel [~marcel@jharkema.xs4all.nl] has joined #linode-xenbeta 02:06 < linbot> New news from forums: A (hopefully useful) Summary of Memory Tweaks in Performance and Tuning 04:32 -!- taupehat [me@taupehat.com] has quit [jupiter.oftc.net kinetic.oftc.net] 04:32 -!- TheFirst [gaveup@your.friendly.neighborhood.hellmouth.info] has quit [jupiter.oftc.net kinetic.oftc.net] 04:32 -!- encode [~encode@blah.i.hate.w1ndo.ws] has quit [jupiter.oftc.net kinetic.oftc.net] 04:32 -!- ElectricElf [~dbharris@electricelf.noc.oftc.net] has quit [jupiter.oftc.net kinetic.oftc.net] 04:33 -!- encode [~encode@blah.i.hate.w1ndo.ws] has joined #linode-xenbeta 04:33 -!- TheFirst [gaveup@your.friendly.neighborhood.hellmouth.info] has joined #linode-xenbeta 04:33 -!- taupehat [me@taupehat.com] has joined #linode-xenbeta 04:33 -!- ElectricElf [~dbharris@electricelf.noc.oftc.net] has joined #linode-xenbeta 06:47 -!- egatenby_ [~egatenby_@itsb149.itsnpt.com] has left #linode-xenbeta [] 09:50 -!- snorp|out is now known as snorp 09:58 -!- brianrumburg [~bri@COX-66-210-194-30.coxinet.net] has joined #linode-xenbeta 09:58 -!- brianrumburg [~bri@COX-66-210-194-30.coxinet.net] has left #linode-xenbeta [] 10:25 -!- pclissold [~pclissold@sassenheim.clissold.nl] has joined #linode-xenbeta 10:29 -!- alnr [~alan@cpe-69-200-85-107.nyc.res.rr.com] has joined #linode-xenbeta 11:39 -!- taupehat [me@taupehat.com] has quit [jupiter.oftc.net kinetic.oftc.net] 11:39 -!- TheFirst [gaveup@your.friendly.neighborhood.hellmouth.info] has quit [jupiter.oftc.net kinetic.oftc.net] 11:39 -!- encode [~encode@blah.i.hate.w1ndo.ws] has quit [jupiter.oftc.net kinetic.oftc.net] 11:39 -!- ElectricElf [~dbharris@electricelf.noc.oftc.net] has quit [jupiter.oftc.net kinetic.oftc.net] 11:40 -!- encode [~encode@blah.i.hate.w1ndo.ws] has joined #linode-xenbeta 11:40 -!- TheFirst [gaveup@your.friendly.neighborhood.hellmouth.info] has joined #linode-xenbeta 11:40 -!- taupehat [me@taupehat.com] has joined #linode-xenbeta 11:40 -!- ElectricElf [~dbharris@electricelf.noc.oftc.net] has joined #linode-xenbeta 11:47 -!- linbot [~supybot@ns.theshore.net] has quit [Ping timeout: 480 seconds] 12:54 -!- jimcooncat [~jim@216-220-225-50.midmaine.com] has joined #linode-xenbeta 12:58 < jimcooncat> how's the beta going? is the only holdup the no-automatic-reboots issue I saw in the forums? 13:00 < snorp> jimcooncat: it appears to be going a little better now 13:01 < jimcooncat> Do you think it will come out of beta when uptimes are better than uml-based? Or will they wait 'til it's near-perfect? 13:02 < snorp> I honestly don't know 13:02 < snorp> but I think it'll be a while still 13:02 < jimcooncat> thanks, I'm anxious for better reliability 13:02 < snorp> ah, stick with uml for now then :) 13:02 < snorp> because it's definitely not better currently 13:02 < jimcooncat> ok, thanks for the info 13:03 < caker> the uptime problem is a BIOS issue on certain hosts, nothing to do with UML 13:04 < jimcooncat> oh, I was mistaken. thanks 13:05 -!- jimcooncat [~jim@216-220-225-50.midmaine.com] has quit [Quit: Leaving] 13:06 < TheFirst> there's still the thrasher issue, no? 13:07 < caker> yup 13:07 < caker> and the stability issues 13:08 -!- linbot [~supybot@ns.theshore.net] has joined #linode-xenbeta 13:09 < TheFirst> hop too ;) 13:09 * TheFirst cracks the whip 13:14 -!- Spads [~crack@dsl081-246-246.sfo1.dsl.speakeasy.net] has quit [Remote host closed the connection] 13:15 -!- Spads [~crack@dsl081-246-246.sfo1.dsl.speakeasy.net] has joined #linode-xenbeta 14:54 < linbot> New news from forums: host56 xenode down in Xen Public Beta 14:55 < taupehat> argh! 14:55 < caker> it is? 14:55 < petert> mine isn't 14:55 < caker> root@host56:/# uptime 14:55 < caker> 14:55:48 up 3 days, 23:18, 2 users, load average: 0.76, 0.68, 0.56 14:55 < taupehat> same 14:56 < taupehat> 11:56:00 up 3 days, 23:06, 1 user, load average: 0.27, 0.20, 0.18 14:56 < petert> looks like a single-node crash 14:56 < caker> yeah 14:58 < taupehat> reckon so 14:59 < linbot> New news from forums: host56 xenode panic in Xen Public Beta 14:59 * caker changed the topic 14:59 < taupehat> ok 15:00 < Spads> !! 15:01 < Spads> caker: what's the story on the Xen I/O runaways? 15:01 < caker> Spads: not much to tell. Thrashers... setting their vxd threads to CFQ's idle class seems to help a bit 15:01 < caker> but nothing like the limiter 15:01 < taupehat> are you going to reimpliment the bitbucket again? 15:01 < Spads> aha 15:01 < caker> looking at the CFQ code, I could probably tweak some of the values (like idle grace period is only 2ms currently) to make it idle heavier, but I haven't messed with it 15:02 < caker> taupehat: either that, or just hold off on Xen until it has its own method to do that 15:02 < caker> I'm not too happy with the stability, either 15:02 < caker> For now, I'm going to focus on putting more UML hosts online that can be moved to Xen with just a reboot 15:02 < taupehat> yeah 15:02 < taupehat> it's certainly kept you busy 15:02 < caker> and do that along with memory upgrades on the existing hosts, regardless of Xen 15:04 < TheFirst> mean so more upgrades to the default plans? ;P 15:04 * TheFirst prays 15:05 < caker> that's the plan 15:05 < caker> that might be put off for two months, but then a big push to get it done 15:05 < TheFirst> that'd be cool...only a few meg shy of having sa working on the 80 now.. :P 15:05 < TheFirst> right 15:05 < TheFirst> hey it may be less mem wise here than other places but the rest makes up for it 15:06 < TheFirst> what about storage? yah now i'm being greedy :P 15:06 < caker> that too 15:06 < TheFirst> hell now all i need is the 'youve got a job' phonecall and this would really be my day 15:09 < taupehat> heh 15:09 < taupehat> hey TheFirst what was it you wanted earlier? 15:10 < TheFirst> gonna ask for ps output for that post i was making 15:12 < taupehat> ? 15:13 < TheFirst> http://www.linode.com/forums/viewtopic.php?t=2243&sid=f972e111f1fb0488c64de21a0ba8aaf3 15:18 < taupehat> TheFirst: so what command you want me to run? 15:18 < taupehat> also, you forgot to mention greylisting for spam control 15:18 < taupehat> postgrey totally rules that world, and has a very light footprint (particularly compared to SA) 15:18 < TheFirst> no...someone else mentioned it, and i note why i didnt 15:18 < taupehat> o 15:19 < TheFirst> ps vax is what all the other output was 15:19 < TheFirst> i was ready to setup greylisting until i saw the need to maintain a whitelist of broken servers 15:19 < taupehat> heh 15:20 < taupehat> that's actually maintained by the package maintainers, at least in the case of postgrey 15:20 < taupehat> and the fraction of broken servers is small enough that it's of little consequence 15:20 < taupehat> I run greylisting at work 15:20 < TheFirst> yes and no...the broken servers are some big names 15:21 < TheFirst> iirc ebay and amazon were on the list 15:21 < taupehat> all sent along with the package 15:21 < taupehat> the default whitelists are pretty extensive 15:21 < taupehat> on the rare occasion I get a phone call about emails being dropped, I get hold of the server admin on the other side and discover that he's running something written in the 90s 15:22 < taupehat> really, most modern MTAs work fine with greylisting 15:22 < TheFirst> like i said in the thread...not something i want to deal with hassle wise 15:22 < taupehat> right 15:22 < taupehat> the thing is 15:22 < taupehat> it hasn't caused any hassles 15:23 < taupehat> quite the opposite 15:23 < taupehat> pre-postgrey (this is my work mailer), we had SA flagging stuff left and right that was wrong while passing really nasty spam right through, and the server load was high. 15:23 < taupehat> users complained, the server was slow, etc. 15:24 < Spads> of course, OS/2 didn't have any native apps, while we do... 15:24 < Spads> mcm 15:24 < Spads> damn I'm clumsy today 15:24 < TheFirst> sure, but given the amount of spam i get on my domains if there were any trouble it'd be more than warrented 15:24 < taupehat> post-greylisting, the server load is healthy, the users are happy, and I get perhaps one false-positive report every three or four months. 15:24 < taupehat> TheFirst: we have 160 non-technical users sending mail from our domain 15:25 < taupehat> many of them have signed up for the free iPod, etc. 15:25 < Spads> hahaha 15:25 < taupehat> you know what I'm talking about =] 15:25 < TheFirst> yah ... thats how it was at my last job...well more around 250 i think but solution to problems there was lock everyone down to work stuff...yah they complained but the amount of problems solved was worth it 15:25 < taupehat> anyhow, greylisting handled a LOT of hassles for me 15:26 < TheFirst> went from like 10 'computer slow' problems a day to 1 a month (from spyware) 15:26 < taupehat> TheFirst: yeah, that's well and good unless the damage is already done, which it is in our case 15:26 < taupehat> I work in K12 - we have a lot of restrictions in place already 15:26 < taupehat> but when I got there, about half the computers didn't even have a/v software installed 15:26 < TheFirst> well it helps when you only need to worry about mail from domains x y and z and can drop everything else 15:26 < taupehat> if you wanted to bypass the filtering proxy, all you had to do was set your IP to a routeable address 15:27 < TheFirst> aside from IT everyone else only needed it for work stuff which was only with certain companies 15:27 < TheFirst> heh people always give me shit cuz i don't bother to update or run a/v at home....my response is so long as you're not a complete dumbass you won't get infected 15:27 < taupehat> yeah, not an option in my environment 15:28 < TheFirst> sure work environments it's not 15:28 < taupehat> but I don't run an a/v at home either 15:28 < taupehat> see, no windows 15:28 < taupehat> no a/v 15:28 < taupehat> =] 15:28 < TheFirst> i have to run windows 15:28 < TheFirst> hardware isn't supported under nix 15:28 < Spads> or your dialysis machine will kill you! 15:28 < taupehat> which hardware? 15:29 < TheFirst> myhd is the big one 15:29 < Spads> he has to run windows or he'll be trapped in our dimension forever! 15:29 < TheFirst> from what people have said the other one (fusion) has support of some kind but i haven't looked 15:29 < taupehat> myhd? 15:29 < TheFirst> atsc tuner 15:29 < taupehat> satellite? 15:29 < TheFirst> ota 15:29 < TheFirst> dvb-s is well supported under linux 15:29 < TheFirst> but utterly useless in the us 15:30 < taupehat> ahh 15:30 < taupehat> I've got a standard wintv card hooked up to rabbit ears 15:30 < taupehat> works fine under linux 15:30 < TheFirst> also video encoding in linux is the suck... 15:30 < taupehat> true 15:30 < taupehat> I use my Mac for that 15:30 < TheFirst> heh i've got rabbit ears too.... just happens to be part of a large outdoor antenna propped on a cd rack 15:31 < TheFirst> and split to the hdtv, and 3 pc tuners 15:31 < taupehat> jeez 15:31 < TheFirst> ? 15:31 * taupehat just did apt-get install mythtv and left it at that 15:31 < taupehat> was much easier =] 15:31 < TheFirst> that's analogue! 15:31 < TheFirst> horrid quality! 15:31 < TheFirst> gimme my hd 15:31 < taupehat> ok 15:32 < taupehat> you have your hd 15:32 < taupehat> I have my bills to pay =P 15:32 < TheFirst> and i'll cut back to 1 tuner when the networks quit putting my shows on at the same times 15:32 < TheFirst> hehe 15:32 < TheFirst> i've got an analogue tuner in my pc too 16:21 -!- pclissold [~pclissold@sassenheim.clissold.nl] has left #linode-xenbeta [] 18:16 < caker> http://lists.xensource.com/archives/html/xen-devel/2006-04/msg01107.html :( 18:17 < caker> I'll raise this question again ... what does it take to run under 64bit besides the kernel? 18:18 < gpd> 64bit libraries 18:18 < caker> and if you don't do that -- kaboom? 18:19 < gpd> no you can have both 64 and 32 with some distros 18:19 < gpd> SuSE has both 18:19 < gpd> I think Ubuntu is working on this 18:19 < caker> I'm thinking of the situation where an end user is moved from 32 to 64bit host, -- they'll need big changes 18:19 < gpd> oh ya - for sure 18:19 < petert> though it can be a huge pain if you only have some libs in 32 or 64 bit 18:20 < caker> that doesn't sound very fun under our environment 18:20 < gpd> and some stuff is not available for 64 - proprietary etc 18:20 < caker> hrmphf 18:20 < gpd> how much RAM on your machines? 18:20 < petert> caker: for this usage, you should be able to push the libs problems down to the users 18:21 < caker> gpd: 16G on these new boxes... like my email says, vanilla (and UML) can see it all, but not Xen 18:21 < caker> I get jipped a GB or so 18:21 < petert> we can always run a 32bit distro image under a 64bit kernel if there's something special we need 18:21 < petert> (or build 32bit libs) 18:21 < caker> petert: oh, that works? ... so technically, even with a 64bit kernel, they can still run their unmodified 32 bit distro? 18:22 < gpd> on ubuntu at present you need a chroot 32bit area 18:22 -!- adinardi [~angelo@tux.csh.rit.edu] has quit [Quit: leaving] 18:22 < gpd> but it will run 18:23 < petert> caker: hmm, sounds like that would need a little testing 18:23 < petert> i thought so, but i don't have first hand experience 18:23 < caker> ok 18:23 < petert> or maybe you'd provide a 32bit kernel too? 18:24 < gpd> http://ubuntuforums.org/showthread.php?t=24575 18:24 < petert> is the kernel that runs on the bare metal the issue? 18:24 < caker> can't mix/match with Xen 18:24 < gpd> if dom0 is 64bit then domUs must be 64 18:24 < caker> if the hypervisor is 64bit, all doms must be 64bit 18:24 < petert> aha, ok 18:25 < gpd> caker: SuSE should work as expected 18:25 < gpd> but i doubt there are many SuSE linoders ;( 18:26 < caker> yeah .. zero 18:26 < petert> i'd probably run one 18:26 < gpd> i would run a 64 bit ubuntu box :) 18:26 < petert> there are a few linoders at novell 18:26 < gpd> the only thing stopping me from running it at home are video codecs 18:26 < gpd> which are not relevant on a server 18:26 < petert> linode usage spread through the ximian offices like a rash one week 18:27 < caker> hah 18:27 < petert> though i'm happy enough running anything really 18:27 < gpd> fedora 4 & 5 should also work with 32/64 18:27 < caker> I *do* need to package up a SuSE template ... I think when I looked into it last year, I got confused by the licensing 18:28 < gpd> opensuse now should be ok 18:29 < gpd> 64 bit server is a little more secure - some might say 18:30 < caker> well, I'm off to buy a label maker that plugs into the computer... bbiab 18:30 < gpd> actually - 18:30 < gpd> https://wiki.ubuntu.com/32bit_and_64bit 18:30 < gpd> sudo apt-get install ia32-libs* 18:31 < gpd> they might have improved matters already 18:57 -!- smohekey [~smohekey@203.98.10.169] has joined #linode-xenbeta 19:02 -!- smohekey [~smohekey@203.98.10.169] has quit [Remote host closed the connection] 20:42 < linbot> New news from forums: Debian/Gentoo ponderings in Performance and Tuning 20:58 -!- snorp is now known as snorp|out 21:24 < caker> huh .. from Ian Pratt (head Xen guy): 21:24 < caker> You can use a 64b kernel on top of a 32b distro just fine. 21:26 < TheFirst> huh 22:38 -!- smohekey [~smohekey@203.98.10.169] has joined #linode-xenbeta 22:51 < gpd> caker: does that work both ways? 22:52 < caker> gpd: doubtful ... 22:53 < caker> I really don't know 22:53 < gpd> curious - but not too hard to test ;) 22:54 < gpd> did you check the BIOS as he suggests? that tooo sounds unlikely 22:54 < gpd> he also says 33b - I suspect he was on the beers 22:54 < gpd> he is a fellow Brit after all 22:55 < caker> I agree 22:55 < caker> (to the bios issue) 22:56 < caker> obviously, that was a typo 23:19 < lucca> hm, any plans to allow custom and/or non-linux xen-nodes? 23:21 < caker> I would if I could manage the filesystems from within Linux 23:22 < TheFirst> not the response i was expecting 23:23 < lucca> hmmm, interesting :p 23:23 < TheFirst> expected something to the effect of: "stable first" 23:23 < lucca> well, much down the road after things are stable, obviously 23:23 < mikegrb> yeah 23:23 < lucca> and clearly no support in the normal sense 23:23 < lucca> hmm, I assume you mean read-write? 23:24 < mikegrb> xen will be quite stable long before freebsd and linux can read/write each other's filesystems 23:24 < lucca> they can Sorta read now 23:24 < mikegrb> sense the developers of each don't crossover very much 23:24 < TheFirst> kind of ironic really 23:25 * gpd notes recent mudslinging by Torvalds to BSD camp :) 23:26 < lucca> heh, not looking for that kind of drama 23:26 < TheFirst> what slinging be this? 23:26 < caker> also, http://www.opensolaris.org/os/community/xen/ 23:26 < caker> I've read they have it booting/working 23:27 < TheFirst> y'know how can all these companies rave about xen....i mean it's not possible that they don't experience the bugs we have....it's like ok xen has potential but there's some pretty big issues still... 23:27 < caker> yup 23:28 < OverlordQ> TheFirst: says that the FreeBSD guys are morons for using CoW 23:28 < TheFirst> and CoW is...? 23:28 < caker> copy on write 23:29 < TheFirst> elaborate for the severly nonfuctional and sleep deprived TheFirst? 23:29 < OverlordQ> "I claim that Mach people (and apparently FreeBSD) are incompetent idiots. Playing games with VM is bad. memory copies are _also_ bad, but quite frankly, memory copies often have _less_ downside than VM games, and bigger caches will only continue to drive that point home." 23:29 < OverlordQ> is what he said 23:29 < caker> start with a read-only whatever (memory, file, disk image) ... something "writes" to it, but it doesn't change the orig.. just saves the changes 23:29 < caker> http://en.wikipedia.org/wiki/Copy-on-write 23:31 < TheFirst> man...things like this (my apparent and complete idiocy) make me wonder how the hell a) i'm ready for a job and b) how in the hell the idiots that are like duh whats an os GET them 23:31 < OverlordQ> linus was basically saying that CoW is worse off then just making a seperate copy in the beginning 23:32 < OverlordQ> due to extra updates and a fault in there somewhere 23:33 < OverlordQ> i wont pretend I understood a single thing they were talking about 23:41 < gpd> The COW concept is also used in virtualization/emulation software such as Bochs, QEMU, and UML for virtual disk storage. 23:42 < gpd> didn't know UML used it ! 23:42 < mikegrb> it can 23:42 < caker> it doesn't work for us, since the backing file needs to be the same size as the resulting disk image 23:42 < gpd> oh 23:42 < gpd> (which is good) as we are not morons! 23:44 < TheFirst> gpd: don't look in the mirror much, eh? 23:44 * TheFirst hides --- Log closed Tue Apr 25 23:59:01 2006