Back to Home / #linode / 2006 / 04 / Prev Day | Next Day
#linode IRC Logs for 2006-04-26

---Logopened Wed Apr 26 00:00:38 2006
00:03<npmr>debian kernel packages have a script in them which runs mkinitrd to generate an initrd at install time
00:03<npmr>the contents of your system's initrd are configured in /etc/mkinitrd
00:07<iggy>I fail to see how booting a 32bit userspace is any different than chroot'ing into one
00:08<npmr>shouldn't be
00:08<iggy>and I chroot into them all the time for testing
00:08<iggy>and compile things without issue
00:09<npmr>the 32-bit libc in your chroot will use the same abi as if it had booted that way
00:09<erikh>caker: perhaps this is pestering, but how are the builds for the new servers coming? :)
00:12<@caker>erikh: done, except for labels
00:12<@caker>erikh: they'll be shipped out tomorrow, and HE has a scheduled install for Thursday
00:12<iggy>DAMN YOU LABEL PRINTER!!!!!!
00:12<@caker>erikh: if all goes well, they'll be online by Friday
00:12<@caker>iggy: no kidding
00:12<erikh>sweet! thanks
00:13<@caker>I refuse to make the labels manually (although, you know that's what I'll be doing tomorrow morning)
00:13<erikh>the life of the system administrator. :)
00:13<@mikegrb>mmm cake
00:13<iggy>the life of cake
00:49|-|DarthClue [] has joined #linode
00:55|-|DarthClue [] has quit [Quit: HydraIRC -> <- The professional IRC Client]
00:55<gpd>so is there any advantage to running this 64bit kernel with 32bit userspace?
00:55<gpd>clearly it is not normal as you cannot download one by normal apt
00:55<gpd>dpkg --unpack wouldn't even touch the 64bit kernel .deb
01:23|-|darkbeholder [] has joined #linode
01:24<gpd>Running the mixed setup on a workstation is not recommended, because iptables, the XFS filesystem, non-free NVidia and ATI binary drivers do currently not support it.
01:25<gpd>gpd@beasty:~$ sudo iptables --list
01:25<gpd>iptables v1.3.1: can't initialize iptables table `filter': Bad file descriptor
01:25<gpd>Perhaps iptables or your kernel needs to be upgraded.
01:25<gpd>caker: or possibly not :( - but then who needs firewalls
01:31|-|linbot [] has quit [Ping timeout: 480 seconds]
01:32|-|linbot [] has joined #linode
01:32|-|adamg [] has joined #linode
01:34<@caker>gpd: what's required to get iptables working? Updated iptables package?
01:34<@caker>gpd: or one that's compiled specifically for 64bit?
01:35<gpd>64bit version it seems - or use ipchains
01:35<gpd>but no 64bit version in most distros
01:35<@caker>ipchains hasn't been in Linux since 2.2, no?
01:35<gpd>that is what i thought
01:38<gpd>chanel becomes digg spy
01:38<gpd>OverlordQ: yeah - i thought that might be an option -
01:39<gpd>but caker would have to work some magic
01:39<gpd>and all this for about 2 Gig... ?
01:39<sargun>I should wrap a present with this stuff:
01:40<gpd>sargun: nice - reminds me of the suicide bunnies
01:40<OverlordQ>yea 'nother page i seen says you'd have to recompile iptables
01:41|-|spr [] has quit [Quit: Spoon!]
01:43<OverlordQ>lol @ bunnies
01:43<OverlordQ>those are always good
01:43|-|HuK0B [~NIKOV@] has joined #linode
01:43|-|konoko [] has quit [Ping timeout: 480 seconds]
01:48<gpd>caker: any thoughts?
01:49<@caker>gpd: yeah, not going to work for us if it requires libs, etc
01:50<gpd>there seem to be patches for iptables
01:53|-|konoko [] has joined #linode
02:02<erikh>caker: IIRC, you can get ipchains support if you compile it in during kernel compiles
02:02<erikh>(in 2.4 at least)
03:05|-|OverlordQ [] has quit [Quit: ]
03:12<neurosis>why does /etc/hosts have this by default in it.... ???
03:12<neurosis>that doesnt look right
03:13[~]neurosis is still grappling with that small slack 10 distro
03:15<neurosis>i mean im no network guru as you all well know... but that sure is an awful long hostname
03:16<neurosis>and i cant imagine it is correct
03:19|-|pclissold [] has joined #linode
03:42|-|kokoko [] has joined #linode
03:49|-|konoko [] has quit [Ping timeout: 480 seconds]
04:35|-|HuK0B [~NIKOV@] has quit [Ping timeout: 480 seconds]
04:43|-|konoko [] has joined #linode
04:49|-|kokoko [] has quit [Ping timeout: 480 seconds]
05:23|-|jekil [] has joined #linode
06:41|-|linville [] has joined #linode
06:56|-|schultmc_ [] has joined #linode
07:09|-|pclissold [] has quit [Quit: Goodbyeeeeeee]
07:13|-|jekil [] has quit [Quit: Leaving]
07:42|-|ElectricElf [] has quit [Read error: Connection reset by peer]
07:52<neurosis>ok i switched to debian
07:52<neurosis>you guys were right really easy to get the pop3 and smtp servers installed
07:53<neurosis>however.... they dont seem to want to send or recieve mail
07:54<neurosis>not recieving it
07:54<neurosis>im going crazy here
07:54<neurosis>plz dont tell me im gonna be forced to use redhat9 again
07:54<neurosis>it used to work fine in redhat
08:07|-|Redgore [~Redgore@] has joined #linode
08:12|-|ElectricElf [] has joined #linode
08:18<iggy>neurosis: that looks like it worked to me
08:22<neurosis>i never received it tho
08:25<sonorous>do the usual, check your logs, check your conf, find out where it's putting the mail etc
08:27<neurosis>i reinstalled fedora core 2
08:27<neurosis>started sendmail service
08:27<neurosis>and blam sent a mail to my yahoo mail account
08:27<neurosis>worked fine
08:28<neurosis>what service do i start to run the pop serveR?
08:31|-|Battousai [] has quit [Ping timeout: 480 seconds]
08:42|-|kokoko [] has joined #linode
08:44|-|Dreamer3 [] has quit [Ping timeout: 480 seconds]
08:45|-|Dreamer3 [] has joined #linode
08:49|-|konoko [] has quit [Ping timeout: 480 seconds]
09:14|-|FireSlash [] has joined #linode
09:55|-|sargun [] has quit [Ping timeout: 480 seconds]
10:04|-|macdan [] has joined #linode
10:12|-|anderiv [] has quit [Quit: Lost terminal]
10:12|-|anderiv [] has joined #linode
10:34|-|jekil [] has joined #linode
10:48|-|jekil [] has quit [Remote host closed the connection]
10:54|-|Battousai [] has joined #linode
11:00|-|FireSlash [] has quit [Read error: Connection reset by peer]
11:01<erikh>what server do you want to run? you should have a few choices.
11:01<erikh>qpopper, courier pop, dovecot pop...
11:01<erikh>if you want some thing simple, qpopper/dovecot would probably be best
11:13<ra>neurosis: qmail (netqmail-1.05) with vpopmail and chkuser works great. best security track-record and easy-to-maintain. downside is you'll need to compile
11:23|-|kokoko [] has quit [Ping timeout: 480 seconds]
11:35|-|konoko [] has joined #linode
11:35|-|spr [] has joined #linode
11:38|-|jekil [] has joined #linode
11:41|-|konoko [] has quit [Quit: reboot]
11:55|-|konoko [] has joined #linode
12:05|-|Redgore [~Redgore@] has quit [Quit: A geek without purpose - | SMDC-Network IRC -]
12:08|-|afv-13 [] has joined #linode
12:11|-|jekil [] has quit [Ping timeout: 480 seconds]
12:13<@caker>wow ...
12:13<@caker>that's RIGHT across the street from me, where I go get goodies
12:16<Spads>If by "goodies" you mean "painkillers"...
12:17<konoko>whats wrong with pain anyway caker ?
12:27<JasonF>caker: no more working on the linode kernels while on the painkillers.
12:27<JasonF>Maybe it'll panic less ;)
12:30|-|adamg [] has quit [Ping timeout: 480 seconds]
12:39|-|konoko [] has quit [Quit: reboot]
12:44|-|Spads [] has quit [Read error: Connection reset by peer]
12:44|-|Spads [] has joined #linode
12:49[~]caker starts making labels
12:54<Spads>>> Speed Dating to Raise Money for The Walk Against Rape on April 29th
12:54<Spads>Amazingly, NOT A PRANK
13:01|-|konoko [] has joined #linode
13:01<afv-13>speed dating, not speed raping
13:04<lucca>labels eh?
13:04<lucca>Paper St. Soap Co.
13:11<Milliways>geez, caker.. the place has been robbed twice in the past 2 months.. maybe you should move to jersey
13:14<@caker>it's Kirb Appeal!
13:15<Milliways>jersey doesn't have much Kirb appeal, from what i gather
13:34<cmantito>jersey. sucks.
13:45<gpd>caker: looks like iptables 32/64bit problems might be fixed in 2.6.17
13:48<gpd>XFS may still be a problem - but you don't support that officially? correct?
13:49<@caker>it's in recent kernels
13:51<gpd>well it doesn't play well with 64Bit kernel & 32bit userland it seems
14:06|-|jekil [] has joined #linode
14:11<ra>kinda neat graph of traffic. not sure about accuracy though (counts only IE with s*ware).
14:13<ra>wonder if linode actually gets around 15,000 - 20,000 hits per day based loosely on that graph.
14:15<@caker>more like around 60k/day
14:16<@caker>we did 1.6mil last month /me flexes
14:16<ra>figures, because of higher percentage of firefox users these days
14:17<ra>that is awesome
14:18<ra>i hope you profit enough to enjoy life to the fullest, but not enough to retire from linode
14:19<ra>hmmm.. do you know what percentage of your visitors are using IE?
14:19<@caker>yeah .. one sec
14:19<@caker>58% "netscape" (I think I made that include all mozilla browsers)
14:19<@caker>29.7% MSIE
14:19<@caker>1% opera
14:19<Spads>ha ha opera
14:20<@caker>the rest are scrapers, etc
14:20<Milliways>29.7% IE.. that speaks volumes as to your clientel
14:20<@caker>that's from this month
14:21<@caker>!calc 75933 + 42790
14:21<linbot>caker: 75,933 + 42,790 = 118,723
14:21<@caker>^-- my hits
14:21<@caker>mike's got about 6k hits
14:22<@caker>although he was out for a week .. he did 14k last month
14:22<Milliways>our work site, 1.3 million hits this month so far, is 54.5% IE
14:22<ra>thanks caker, this info is very helpful in testing out my formula to convert alexa stats into actual hits
14:22<@caker>we're at 1.34 mil hits this month, too
14:23<TheFirst>oo when'd the annoying competiter split?
14:25<ra>here is what i'm trying... slashdot around 7/2004 reported 2 million daily hits. they ranked about 1,000 at that time. so if ranks 15 (today) 22 (avg), i do 0.022 * 2 mil to get linodes approx daily hits of 30,000 - 74,000.
14:27<@caker>for this month, the worst/best days were 35k/105k hits, with 59k average, or so
14:27<ra>the reason i'm doing this is to approximate traffic i might get if a news article was posted about my site. i'm using the spikes generated by the site to get an estimate about CTR
14:28<ra>i'm seeing produce about 10% - 20% CTR if a product gets mentioned in their news. so i'm trying to anticipate traffic. i'd hate to think what slashdot would do
14:28<@caker>just make sure your apache limits are set, and you'll be fine
14:30<ra>i think i represent the 1% opera :)
14:38|-|Redgore [~Redgore@] has joined #linode
14:40<ra>can you tell how many of's hits are from unique users?
15:09|-|Redgore [~Redgore@] has quit [Quit: A geek without purpose - | SMDC-Network IRC -]
15:25<gpd>ra: have you tried openload ?
15:25<ra>i've used siege and ab (apachebench)
15:26|-|neurosis [] has quit [Read error: Connection reset by peer]
15:43<@caker>servers have been shipped!!
15:44<sonorous>what spec? :)
15:45<gpd>caker: did you decide to suck up the memory loss?
15:45<@caker>same as host57 .. dual 2.2Ghz AMD, 16G RAM, 3ware 9550SX 500GB Raid1
15:45<@caker>gpd: these won't be xen boxes
15:46<taupehat>not this week
15:46<gpd>caker: so how do you convert a UML box to a Xen box? do you ship them back?
15:46<@caker>gpd: I reboot it
15:46<@caker>gpd: all new hosts are ready to go, either UML or Xen
15:47<@caker>host1-55 require them to be empty, so I can repartition them (LVM)
15:47<@caker>but I can do that with netboot+initrd
15:48<taupehat>netboot ftw
15:48[~]taupehat runs LTSP servers, netboot is a primary technology there
15:49<gpd>so - now that is done... time for Xen i/o magic
15:50<@caker>I'm so burnt out on Xen, but yeah ...
15:50[~]gpd cracks the whip
15:50<taupehat>caker: just use vmware =P
15:50|-|Milliways changed nick to kriby
15:50<taupehat>get like 6 nodes per server
15:50<@caker>I'm going to look at hacking cfq -- either a new prio "class" or modifying the existing one
15:51<gpd>nobody at xen doing this in the near future then :(
15:51<@caker>not that I'm aware of
15:51<@caker>XenEnterprise has some text that eludes to it, but it's not clear
15:52<taupehat>you'd think they would care about xen hosts not being overwhelmed by one machine
15:52<@caker>well, they don't use it in an environment like ours
15:52<gpd>XENterprise <-- marketing baby
15:53<taupehat>es todos esta es
15:53<@caker>"Per guest resource guarantees for CPU, memory and I/O" <-- I/O can refer to a lot of things
15:53<@caker>network IO
15:53[~]caker shrugs
15:54<@caker>I signed up for their beta .. no word yet
15:54<gpd>can you pay someone to do it >)
15:54<@caker>I don't need to, plenty of options
15:54<@caker>plus, it's not that big of a deal IMO
15:54<@caker>UML is stable, fast (getting faster) and proven
15:57<gpd>xen is dead - long live uml?
15:57<gpd>at least until they fix PAE and IO ?
15:58[~]gpd strokes UML linode, cackles manically
16:08|-|linville [] has quit [Quit: Leaving]
16:13[~]caker builds 2.6.17-rc2-um
16:14[~]kriby considers actually upgrading to 2.6 and some point, and decides to delay it a bit longer.
16:14<TheFirst>i went to 2.6 on my homebox...don't remember what but had lots of problems with it
16:15<kriby>i've been running 2.6 at home for a long time.. just haven't felt like messing with something that's working on the 'node
16:17<afv-13>i was using 2.6 at home in the rc stages thanks to kernel alsa and being too much of a noob for oss
16:30<linbot>New news from forums: Kernel: with NPTL/TLS support in Announcements <>
16:42<linbot>New news from forums: Kernel: 2.6.17-rc2-linode19 (with NPTL/TLS support) in Announcements <>
16:45<linbot>gpd: The latest stable kernel is; the latest snapshot of the stable kernel is 2.6.17-rc2-git8; the latest beta kernel is 2.6.17-rc1-mm3.
16:51|-|weasel [] has quit [Read error: Operation timed out]
16:51|-|D[a]rkbeholder [] has joined #linode
16:53|-|Netsplit <-> quits: darkbeholder, cow, Jay-dargo, tsi, iggy, sprouse, flatronf700B, gsf, AndyHat, besonen_, (+2 more, use /NETSPLIT to show all of them)
16:53|-|Netsplit over, joins: D[a]rkbeholder
16:53|-|Netsplit over, joins: darkbeholder
16:53|-|Netsplit over, joins: flatronf700B
16:53|-|Netsplit over, joins: besonen_
16:53|-|Netsplit over, joins: Jay-dargo
16:53|-|Netsplit over, joins: tsi
16:53|-|Netsplit over, joins: AndyHat
16:53|-|Netsplit over, joins: gsf
16:53|-|Netsplit over, joins: sprouse
16:53|-|Netsplit over, joins: schultmc
16:53|-|Netsplit over, joins: iggy
16:53|-|Netsplit over, joins: cow
16:54|-|darkbeholder [] has quit [Read error: Operation timed out]
16:56|-|weasel [] has joined #linode
17:11|-|adamg [] has joined #linode
17:45[~]gpd thrashes oO
17:46<gpd>seems to be presenting me with the opportunity to reboot into 17-rc2
18:01|-|spr [] has quit [Quit: "Ah, the bounce has gone from his bungie"]
18:16|-|macdan [] has quit [Quit: Got Itch?]
19:00|-|afv-13 [] has quit [Ping timeout: 480 seconds]
19:04|-|linbot [] has quit [Ping timeout: 480 seconds]
19:12|-|linbot [] has joined #linode
19:14<@caker>hrm .. what else to try with this box that won't POST (not host39)
19:14<@caker>drives: removed, pci cards: removed, ram: bisected
19:15<gpd>unplug the power for 1 minute ?
19:15<@caker>it's been unplugged for a few days...
19:15<gpd>do a little dance?
19:15<@caker>tried that already
19:15<gpd>threaten it with violence? (i'm not helping)
19:16<gpd>does it beep with no memory?
19:16<@caker>No :(
19:16<gpd>then you may have a high tech door stop
19:17<gpd>you could try a different PSU...
19:17<@caker>wth happened to it? It's been in the server closet unused for over a year
19:17<gpd>reseat the cpu?
19:17<@caker>I guess that's next ...
19:17<gpd>not much left
19:17<@caker>actually, I have another
19:18<@caker>it's dual proc MB but I only ever had one of them in it
19:21|-|adamg [] has quit [Ping timeout: 480 seconds]
19:22<gpd>what FS should I use for a 400G drive for rw on OSX|Win|Linux - Fat32 seems painful
19:22<gpd>i use ext3 driver for windows but it is a bit sketchy
19:24<@caker>I think you're stuck with fat32, unless you use something that can read Mac drives on Windows (MacDrive?)
19:24<@caker>I dunno what HFS support is like under Linux
19:25<gpd>I think HFS is ok but HFS+ or whatever is in Tiger is ro / unsupported
19:25<@caker>there are also file naming issues to be aware of
19:25<gpd>but i may be wrong
19:26<@caker>oh, only read in XP/2000
19:26<@caker>r/w in Vista
19:26<@caker>r/w in MacOSX and Linux 2.6
19:27<gpd>oh - shiny
19:27<gpd>i presume XFS / JFS and all those are not supported outside Linux
19:27<Spads> ha ha
19:28<gpd>hold the phone... udf supports BSD and Linux...
19:28<gpd>spot the obvious Xen connection
19:28<gpd>wait... read only :(
19:28<@caker>yeah, but BSD needs to boot from its native filesystem type, or some other limitation
19:28<@caker>I saw that as a possible work-around, too
19:29<@caker>BSD can do ext2/3, too, I believe (but, not boot from)
19:29<@caker>UFS is its native, I believe
19:29[~]caker needs food
19:30|-|spr [] has joined #linode
19:37<gpd>Some say that Windows NTFS does not really offer a journaled file system. The current version of NTFS (the predominant file system within Windows 2000 and XP), does not handle full-fledged journaling; change-journal logs note alterations to files but cant provide enough information to reverse them.
19:37<gpd>wow - I thought NTFS was supposed to be 'for teh win'
19:37<gpd>what a delusional world of MS marketing i must live in
19:40|-|schultmc_ [] has quit [Quit: Client exiting]
19:48|-|FireSlash [] has joined #linode
19:56|-|jekil [] has quit [Quit: Leaving]
20:06|-|Marcel [] has quit [Read error: Connection reset by peer]
20:19|-|D[a]rkbeholder [] has quit [Ping timeout: 480 seconds]
20:45|-|Dreamer3 [] has quit [Ping timeout: 480 seconds]
20:51|-|OvrLrd-Q [] has joined #linode
20:52|-|Dreamer3 [] has joined #linode
20:57|-|vitrum [] has joined #linode
21:00|-|darkbeholder [] has joined #linode
21:27<OvrLrd-Q>aaaah, you scared me
22:07<@caker>fo0bar: I played with the PPC6700 .. cool stuff
22:08[~]caker wants
22:09<erikh>ext2/3 on BSD is not ideal
22:09<erikh>nor stable
22:09<erikh>the manpage basically says, "if you don't like how unstable it is, fix it yourself" or something
22:09<erikh>lots of problems with the GPL keeping BSD devs from just taking the code and using it
22:10<@caker>figures ... 2.6.17-rc3 is out
22:10<@caker>the HTC thing
22:11<@caker>DIDN'T YOU GET MY EMAIL?
22:11<@caker>I demand a response!!
22:11<fo0bar>yes, and I replied to it, I just didn't know of sprint's marketing name
22:11[~]caker looks
22:11<fo0bar>do you plan on getting it before I come down?
22:11<@caker>ahh, here it is
22:12<@caker>I dunno
22:12<@caker>I figure I could get it, then you could show me all the hacks
22:12<@caker>(that's if I decide I need it)
22:12<fo0bar>heh, I haven't done much "hacking"
22:12<fo0bar>just installed a few apps
22:12<@caker>just read your reply .. so, not liking it now?
22:13<fo0bar>I've got a ton of MP3s on there now, as well as some videos
22:13<@caker>or you ARE liking it
22:13<fo0bar>supposedly it's really easy to unlock the phone, but since I'm with cingular for another 2 years, there's no need
22:14<@caker>they tuck the stylus away in the antenna on the Apache
22:15<@caker>although on the test model, it was just falling out
22:15<fo0bar>the apache is the one with the full-sized stylus, right?
22:15<@caker>also, the test model was running version 5.0
22:15<@caker>It was pretty big, yeah
22:15<@caker>I jumped on CGI::IRC and it worked fine (IE)
22:15<fo0bar>the wizard has this telescoping stylus, which is neat (something for me to futz with)
22:16<fo0bar>IE sucks on there
22:16<fo0bar>get opera
22:16<fo0bar>if you buy it :)
22:16<fo0bar>"look, over there!" [installs opera on demo]
22:16<@caker>what did you lay out for yours? And does cingular give credits towards a new phone?
22:17<@caker>Sprint gives $75/year towrads a new phone, I believe
22:17<fo0bar>$300 with 2 year contract extension, retail is $500 IIRC
22:18<@caker>$599 regular price for the 6700 ..
22:18[~]caker gasps
22:18<fo0bar>but you can just buy them retail. my boss bought one for the CEO at retail (who has verizon), hacked it to remove the provider lock, and set it up for verizon
22:18<@caker>They're on ebay for around $430
22:20<fo0bar>oh, if you buy one, get a screen protector sheet for it first thing
22:20<fo0bar>I had unusually good luck with the generic sheets that I just trimmed down
22:21<gpd>not tempted by a super micro pc then? eg. OQO?
22:21<fo0bar>gpd: I saw it a few years ago, wasn't impressed
22:22<fo0bar>speaking of supermicro, Pentium D 930 1U supermicro system == kickass
22:23<fo0bar>we have a project at work where we need about 25 windows test machines, so I got that + vmware server
22:23<fo0bar>on a linux host
22:23<@caker>huh .. all 25 test VMs on that box, eh?
22:23<fo0bar>well, not at the same time
22:23<fo0bar>6-8 concurrent workstations are fine though
22:23<fo0bar>over 10 is pushing it
22:24<@caker>due to the workload or VMWare?
22:24<fo0bar>the workload in general
22:24<fo0bar>vmware will let you start as many as you want
22:24<fo0bar>remember, this is windows :)
22:25<fo0bar>I wanted to try a xen installation, since this proc is VT enabled, but we didn't have time
22:26<fo0bar>vmware server is kickass though. you can install it on a headless linux server, and download a client for windows or linux, and it basically acts like a full copy of vmware workstation
22:27<fo0bar>multiple people can interact with a vm at once
22:27<fo0bar>the biggest downside of server is no sound support
22:27<@caker>fo0bar: what was the thing to watch out for? ROM version?
22:27<fo0bar>watch out for?
22:27<@caker>something about an old software or firmware version on these phones?
22:28<fo0bar>I don't remember anything about that
22:28<gpd>have any palm 'access linux platform' devices come out yet?
22:28<gpd>or is palm dead in the water
22:29<fo0bar>I don't like the palm interface (look at me, I chose windows mobile over palm!), but they seem to keep pumping out treos
22:30<@caker>fo0bar: Weren't you saying that friends of yours have the older version (of software?) and it does weird stuff on them?
22:30<@caker>maybe I read this elsewhere
22:30<gpd>looks like all that stuff is a ways away
22:31<fo0bar>caker: possibly, I don't remember anything like that.... though the "ROM" is just flash; you can upgrade it to the latest that sprint/cingular/etc offers
22:31<fo0bar>I was saying that the apache is slightly inferior to the wizard
22:31<fo0bar>expect 4 hours of video instead of 6, for example
22:32<fo0bar>due to smaller battery
22:32<fo0bar>I'm not sure about talk time though
22:32<fo0bar>caker: BTW, we're moving that 9550SX system into production, FINALLY
22:33<fo0bar>all the problems have been worked out
22:33<@caker>I shipped three more 9550SX systems this afternoon, ordered three more today
22:33<@caker>still working off my box of 10 9550SX cards and BBUs
22:33<@caker>oh? Did you dismiss that corruption problem down to the CPU?
22:33<fo0bar>the first problem was for some reason, compiling the new 3w-9xxx module on rhel4 and replacing it on top of the stock kernel would produce data corruption similar to bad memory under high load (bits flipped, etc)
22:34<fo0bar>we replaced stuff randomly, at one point replacing the mobo/chassis
22:34<fo0bar>at that point, one of the CPUs must have been seated incorrectly
22:35<fo0bar>which trigged another set of problems (filesystem crapping itself and going read-only under high load)
22:35<fo0bar>after playing with the CPUs, I fixed problem #2, but we still needed RHEL4's stock kernel for Oracle
22:36<fo0bar>for some reason (I'm still not sure why), downloading the kernel SRPM, replacing 3w-9xxx.[ch] with new versions from 3ware, and recompiling works
22:36<fo0bar>as opposed to compiling a new module against the kernel headers
22:36<@caker>doesn't sound like much fun
22:37<fo0bar>I've been nursing this damn system since december
22:37<fo0bar>but it saved us $25k
22:37<fo0bar>over a $30k HP "solution"
22:37<@caker>Enterprise == $$$$$$
22:38<fo0bar>yeah, it doesn't help this is for another department, which is super afraid of doing anything themselves. hence hp, oracle, red hat, etc
22:38<@caker>I watched Healthstream dump millions on "solutions" into the ether
22:38<@caker>!stock HSTM
22:38<linbot>caker: The current price of HSTM is 4.5099, as of 3:59pm EST. A change of -0.3701 from the last business day.
22:38<@caker>amazing .. they're up about 200%
22:40<fo0bar>this company is still rather divided into 2 camps, from the 2 companies pre-merger in 2002. one company was suit-and-tie techs with "enterprise" scattered everywhere, and the other was debian, off-the-shelf hardware, and massive amounts of duct tape
22:40<fo0bar>each side has moved a bit toward the center, but not totally
22:41<fo0bar>hardware-wise, supermicro is a nice medium, since it's not "enterprise", but not really off-the-shelf either
22:43<gpd>caker: did anyone boot into the 17-rc2 kernel?
22:44<@caker>gpd: dunno
22:46<fo0bar>if I did one great thing for this company, it was introducing sata and 3ware
22:46<fo0bar>"whoah, we can have hotswap disks for less than the price of a BMW?"
22:47<npmr>ha ha
22:48|-|cook [LinodeJava@] has joined #linode
22:49<gpd>stupid q: does regular sata hotswap or is that a feature of 3ware cages?
22:50<fo0bar>the sata standard allows for hotswap, but the drivers must support it
22:50<fo0bar>many windows drivers will, but most sata controllers are not "true" raid
22:51<fo0bar>ie, it's "hardware assisted software raid"
22:51<npmr>so does that mean all sata hardware supports hotswap?
22:51<fo0bar>and the linux equivalents usually AREN'T hotswap-aware
22:52<fo0bar>3ware is in a unique position that is true hardware raid, and the linux drivers "just work"
22:54<fo0bar>strange answer, but that's the best I can give. the hardware specs of sata are built for hotswap (IE, ground is the farthest-out pin on the connector, etc), but it really depends on the drivers
22:55<gpd>talking of sata - do you think it is worth shifting my desktop's base linux install from an old PATA (26MB/sec) to my new shiny 400GB SATA (60MB/sec)?
22:55<fo0bar>and yes, if you have a compatible controller and drivers, every hard drive will work with hotswap, every cable will work with hotswap, etc
22:56<fo0bar>gpd: in short, yes
22:56<fo0bar>if you can afford it :)
22:56<gpd>i am in the process of doing it - so i like your answer :)
22:57<fo0bar>although I rarely see 60MB/s on a single SATA drive, it's usually closer to 55MB/s
22:57<fo0bar>but still, plenty fast
22:57<fo0bar>Timing buffered disk reads: 164 MB in 3.01 seconds = 54.48 MB/sec
22:58<fo0bar>in my tests, bonnie++ block read results are almost identical to hdparm results
22:59|-|VS_ChanLog [] has left #linode [Rotating Logs]
22:59|-|VS_ChanLog [] has joined #linode
22:59<gpd>what would you use as a general test of a system speed?
22:59<gpd>i tried 'contest'
23:00<fo0bar>I used to use "bonnie++ -n 0", and "block read" figures to compare
23:00<fo0bar>but like I said, "hdparm -Tt" is almost identical, and takes 3 seconds instead of 10 minutes
23:00<fo0bar>but bonnie++ will give you more metrics
23:01<fo0bar>oh, system speed? not disk speed?
23:01<gpd>i liked contest 'cos it did a bit of disk and a bit of cpu to give a general idea
23:01<gpd>yeah - kinda overall handywavy 'speed' that could be used to compare two systems or two settings etc
23:02<gpd>i was trying to compare the 64bit kernel 32bit userland with the 32bit kernel/userland
23:02<fo0bar>a single kernel "make defconfig && time make bzImage" is a nice overall test
23:02<fo0bar>but that's more weighted toward CPU
23:02<gpd>that is pretty much what contest does
23:03<fo0bar>BTW, if you just want to beat the crap out of a system, use this:
23:03<fo0bar>I wrote it, and it works well as a "burn in"
23:03<gpd>lol - hammer time
23:04<fo0bar>generally I run 1-2 copies per CPU
23:06|-|cook [LinodeJava@] has quit [Read error: Connection reset by peer]
23:06<gpd>how does that compare with 'stress'
23:06<gpd>or 'cpuburn'
23:07<fo0bar>cpuburn just pegs the CPU (which is something you want to do during burn-in as well)
23:07<fo0bar>I haven't used stress
23:07<fo0bar>but looking at it now
23:07<gpd>wow - there is also 'spew' and 'crashme'
23:10<fo0bar>stress looks interesting
23:11<gpd>yeah - i identified a 'hardware problem' with that
23:15<gpd>is there an obvious reason why an ssh tunnel should 'timeout'
23:16<gpd>or otherwise disconnect for no reason?
23:18<iggy>I would guess firewall somewhere inbetween
23:19<gpd>any way to stop it from doing it?
23:20<iggy>remove the firewall? send "keepalives"? up the limit in the firewall?
23:22<fo0bar>top - 22:22:35 up 6 days, 6:48, 8 users, load average: 29.16, 14.02, 8.83
23:23<gpd>Add ClientAliveInterval 300 to /etc/ssh/sshd_config ?
23:24<iggy>that was purely speculation
23:28<fo0bar>gpd: thanks for showing me stress
23:29<fo0bar>stress -v -c 8 -i 4 -m 8 --vm-bytes 1G -d 4 --hdd-bytes 2G <-- brings a dual xeon HT 3.4ghz/8GB memory/6x400GB to its knees (in a good way :)
23:30<gpd>yeah it certainly lives up to its name
23:32|-|baCkEr [] has joined #linode
23:32<@caker>access denied
23:36<gpd>ooh - sagiable
23:37<gpd>Dmidecode was first written by Alan Cox... say no more
23:56|-|FireSlash [] has quit [Ping timeout: 480 seconds]
---Logclosed Thu Apr 27 00:00:35 2006