Back to Home / #xen / 2005 / 04 / Prev Day | Next Day
#xen IRC Logs for 2005-04-30

---Logopened Sat Apr 30 00:00:03 2005
00:50--- <<-- Surferdude [] has quit (Quit: HydraIRC -> <- IRC with a difference)
01:54--- <<-- drbyte [] has quit (Quit: Leaving)
03:37--- <<-- DEac- [] has quit (Ping timeout: 480 seconds)
03:50--- ---> DEac- [] has joined #xen
05:32knewt aaaarrrggghhhh. basic dd use of my block device works fine, but put an ext2 on it and fill until it runs out of space and ext2 falls over with errors. so it must be going down a different, buggy, path in the second case. gah.
08:01--- ---> rusty [] has joined #xen
08:30--- <<-- rusty [] has quit (Ping timeout: 480 seconds)
10:20--- ---> hollis [] has joined #xen
10:22--- ---> knewt_ [] has joined #xen
10:25--- <<-- knewt [] has quit (Killed (NickServ command used by knewt_))
10:25--- User: *** knewt_ is now known as knewt
10:53--- ---> bunoc [] has joined #xen
10:54bunoc hello. anybody pls tell me can i convert my domain configuration file to .sxp format ?
10:55bunoc is possible, any tool to do that?
10:55knewt as long as your configuration file isn't dynamic it's easy. just add the -n parameter to xm create, like it says in the help
10:59bunoc knewt: then how can i ask xm to use that created file? xm create -c <created-sxp> ?
11:00knewt -F instead of -f, again mentioned in the help
11:01knewt (-f being optional, so you might not have been using it)
11:04bunoc unfortunately xm help is so long, and scroll down my screen
11:05knewt actually, it'd be "xm help create", and you just pipe it though less anyway
11:05knewt (or xm create -h)
11:06bunoc knewt: that helps. thank you
11:10riel you can always scroll back up your screen ;)
11:11knewt yay, my driver appears to be working (that bit of the functionality actually coded so far, anyway). i think on monday i can safely ask for some victims^Wfools^Wguineapigs^Wvolunteers
11:11bunoc riel: rite, but i cannot scroll up too far. there is a limition
11:12riel on such occasions, I ssh into the computer from another system
11:12riel and scroll up inside my xterm
11:12riel in fact - my test system doesn't even have a monitor or a keyboard ;)
11:12riel just a serial console
11:13knewt i need to get round to reactivating scrollback mode on my console server. i turned it off by mistake and it's really annoying sometimes
11:14knewt thing is, i have to log into the console server physically to do so, since i deactivated everything but console connections over the network
11:14bunoc riel: i am at aterm. still some limition on scrollback
11:15bunoc ok get to bed now. thank you all
11:15--- <--- bunoc [] has left #xen ()
11:37--- ---> Surferdude [] has joined #xen
11:37Surferdude anyone here? I need help with brctl
11:44riel that's a bit vague ;)
11:44Surferdude Sorry... I compiled brclt
11:45Surferdude ./configure and make
11:45Surferdude everything went ok... but xen's still says it isnt installed
11:46riel did you install it after compiling ?
11:46Surferdude yes... after compiling brctl i tried running
11:47Surferdude the command brctl dosnt work, but you can access brctl via the directory
11:47Surferdude and run it from there
11:48riel there's your problem
11:48riel the command brctl should just work
11:48riel otherwise you'll get a nasty surprise when starting up xend
11:49Surferdude So how do i fix it? put an alias in bash_profile?
11:49knewt Surferdude: you did "make install" after the "make", right?
11:50Surferdude no... guess i should
11:51Surferdude [root@plain brctl]# make install
11:51Surferdude mkdir -p /usr/local/sbin
11:51Surferdude /usr/bin/install -c -m 755 brctl /usr/local/sbin
11:51Surferdude [root@plain brctl]#
11:51Surferdude :/
11:51Surferdude brb
11:51knewt ok, so it's installed it into /usr/local/sbin now
11:51knewt riel: fancy being one of my volunteers sometime next week?
11:51riel volunteering for what ?
11:52knewt testing out the driver i've been writing
11:52riel what's it for?
11:52knewt copy-on-write devicemapper target
11:53riel nice
11:53riel that sounds like fun
11:54knewt it's still very early alpha status (several bits of planned functionality not coded yet for one thing), but it /appears/ to be working ok now
11:56knewt i had a hateful moment at one point where i gave an invalid value to another bit of kernel code by mistake, and instead of validating it it just accepted it and then shortly afterwards crashed
12:16xai eeyores birthay today!
13:04--- <<-- hollis [] has quit (Ping timeout: 480 seconds)
13:54--- <<-- knewt [] has quit (Quit: leaving)
13:56--- <<-- muli [] has quit (Ping timeout: 480 seconds)
14:18--- ---> knewt [] has joined #xen
14:27--- ---> _MarkW [] has joined #xen
14:58_MarkW knewt: *ping*
15:05knewt _MarkW: 'lo
15:05_MarkW hey, how's it goin?
15:06_MarkW any luck with the CoW driver?
15:07knewt good. i appear to have working basic functionality up and working :)
15:07_MarkW wicked :-)
15:07_MarkW booted a domain off it yet?
15:08knewt nope, but shortly after i got the full code written i /did/ move from testing it directly on my laptop to inside of a domU on one of my servers. got tired of having to reboot my laptop quite so often :)
15:09_MarkW heh :-) should speed things up.
15:09knewt for the moment i've been testing on top of loopback devices inside the domU
15:09_MarkW OK, sounds like a good plan - much faster rebooting a domU than a whole machine.
15:09_MarkW (plus Kip's domain coredump code is available in the unstable tree now)
15:09knewt i've been editing the source on my laptop, building an updated domU kernel, scping it to the server and booting up the domU
15:10knewt haven't actually moved over to -unstable yet tbh
15:10Surferdude anyone know how to switch from lilo to grub?
15:10_MarkW knewt: no, probably sensible ;-)
15:11_MarkW Surferdude: first step - make sure you have a means to boot your machine and fix things if you toast your MBR ;-)
15:11Surferdude i have a kvm... hopefully that should be enough
15:12_MarkW do you have physical access to this machine if necessary?
15:12Surferdude guess lilo -u /dev/hda should do it
15:12knewt my main moment of hate was when i tracked down the reason for an annoying, very repeatable oops. another bit of the kernel that i was calling into doesn't bother to validate incoming parameters, leading to the crash happening within an entirely different kernel thread from my code
15:13_MarkW knewt: lots of moments of hate to be had when hacking the kernel ;-)
15:13Surferdude ok... lilo is removed
15:13Surferdude how do I get grub on?
15:13_MarkW Surferdude: you're moving quickly ;-)
15:14Surferdude =P
15:14_MarkW have you installed the grub packages for your distro?
15:14Surferdude yup
15:14Surferdude typing grub at an ssh prompt brings up the grub menu
15:14_MarkW there's a program called "grub-install" which may suffice (read the warnings in the man page) or you can do it manually from the grub prompt.
15:15knewt i was passing a sector count, and had forgotten to check first that it was actually > 0. the other code just assumed that the count was at least 1, and it led to falling off the end of a linked list, trying to dereference an invalid pointer
15:15_MarkW knewt: eeek! that's rather non-obvious :-(
15:15Surferdude [root@plain boot]# grub-install /dev/hda
15:15Surferdude Installation finished. No error reported.
15:15Surferdude This is the contents of the device map /boot/grub/
15:15Surferdude Check if this is correct or not. If any of the lines is incorrect,
15:15Surferdude fix it and re-run the script `grub-install'.
15:15Surferdude # this device map was generated by anaconda
15:15Surferdude (fd0) /dev/fd0
15:15Surferdude (hd0) /dev/hda
15:15Surferdude [root@plain boot]#
15:15Surferdude I hope that works
15:16knewt hehe. grub-install works great for basic stuff. of course, i've had great fun installing grub on my systems other than the laptop *g*
15:16_MarkW Surferdude: looks sane. You'll need to write a menu.lst file.
15:17Surferdude looks like one was generated
15:17Surferdude and all the entrys look good
15:17knewt my favourite was when i installed linux on my dad's system for him in a dual-boot config. the bios on this system thinks it's helpful to change the "bios order" of drives depending on what the boot device was
15:17_MarkW knewt: what are you doing for storing block remappings? one of the problems I'm aware of with LVM is it's painful memory usage due to storing all exceptions in RAM.
15:18_MarkW knewt: that's just weird! bios writers are twisted human beings!
15:18_MarkW Surferdude: if you're happy with it (and have a way to get at the machine and fix it) then it's probably time to reboot.
15:18Surferdude just did
15:18_MarkW it work?
15:18* Surferdude watches the kvm
15:19* _MarkW crosses his fingers
15:19Surferdude grub menu is showing
15:19* Surferdude presses enter and crosses his fingers
15:19_MarkW woohoo! :-)
15:19Surferdude Looking good...
15:20Surferdude Yup... at a login prompt on the kvm
15:21Surferdude Time to edit menu.lst and boot into xen (hopefully)
15:21_MarkW wicked! Xen next :-)
15:21_MarkW Surferdude: you using Debian by any chance?
15:22Surferdude nope... DC refused to install it
15:22Surferdude RHEL only... fighres
15:22Surferdude *figurez
15:22Surferdude **figures
15:24Surferdude Ok... so i need to configure my grub.conf
15:24_MarkW I thought RHEL used grub by default... (I guess it's an install-time option)
15:24Surferdude Dont ask... EV1 is weird
15:24_MarkW Yup. There's some guidelines in the manual.
15:25_MarkW Should help you get Xen up and running, then you can start installing domains!
15:25Surferdude which version of vmlinuz should i use?
15:26_MarkW You need to use the vmlinuz-foo-xen0
15:27_MarkW you can run any domain using a xen0 kernel. xenU is the same but without drivers for real hardware (hence smaller) - it can only run in unpriv domains.
15:27Surferdude kernel /vmlinuz-2.6-xen0
15:28Surferdude now for root
15:28_MarkW hang on...
15:28_MarkW kernel xen.gz
15:28Surferdude k
15:28_MarkW sorry: kernel /xen.gz
15:28_MarkW module /vmlinuz-2.6.whatever-xen0
15:28_MarkW k?
15:29Surferdude ok
15:29Surferdude and what do i do for root
15:29_MarkW set it the same as for your other grub entries.
15:29Surferdude Ok
15:29_MarkW make sure the path to xen.gz and vmlinuz is the same as the path for the other grub entries
15:30_MarkW what flavour of RHEL is this? you use udev?
15:30Surferdude i think its version three (taroon)
15:31Surferdude title XenLinux 2.0.5
15:31Surferdude root (hd0,0)
15:31Surferdude kernel /xen.gz
15:31Surferdude module /vmlinux-2.6-xen0
15:31Surferdude good?
15:31knewt _MarkW: quick calculation says that with the default chunk size, a 2GB cow mapping will take up about 164Kb, increasing as required during high concurrency cow-ing
15:32knewt i think i've got the calculations correct
15:32_MarkW Surferdude: yes but you need to add some arguments to the xen and vmlinux lines
15:32Surferdude like the ro and root=/dev/sda
15:32_MarkW have a look at the user manual for some examples
15:32_MarkW yup.
15:32_MarkW knewt: OK, that sounds not too bad :-)
15:32_MarkW what does the on-disk layout look like?
15:33Surferdude title XenLinux 2.0.5
15:34Surferdude root (hd0,0)
15:34Surferdude kernel /xen.gz ro root=/dev/sda
15:34Surferdude module /vmlinux-2.6-xen0
15:35knewt i'm going to create a mode with a small header to the bitmap, but the basic version i've started with is a pure bitmap, nothing else. you can use completely seperate bitmap and cow devices if you want (and i have been during testing)
15:36_MarkW Surferdude: nope. ro and root are arguments that Linux is interested in
15:36_MarkW so they go on the Linux line. there are different arguments for Xen itself.
15:36Surferdude Yeah, i was just looking at the manual
15:37knewt _MarkW: i've also managed to keep concurrency within the in-memory structures about as high as i think is possible :)
15:37Surferdude title XenLinux 2.0.5
15:37Surferdude root (hd0,0)
15:37Surferdude kernel /xen.gz dom0_mem=131072
15:37Surferdude module /vmlinux-2.6-xen0 root=/dev/sda ro
15:37Surferdude I think i got it this time
15:37_MarkW knewt: that's good to hear
15:38_MarkW Surferdude: I thought your root was on hda?
15:38knewt i'll try and come up with a nice ascii-diagrammed text file describing the algorithm if you're interested?
15:38Surferdude I think so...
15:39Surferdude and all the other entrys are root=LABEL=/
15:40Surferdude maybe i should try that?
15:40_MarkW knewt: only if you have time :-)
15:40_MarkW the main problems with previous solutions were memory usage and the on-disk layout
15:41_MarkW and performance of course ;-)
15:43knewt once a particular chunk has been cow'ed it's virtually no overhead above a simple linear mapping for writing (a single bit-test). and reading is always just basicaly a single bit-test over a linear
15:43_MarkW a couple of other things that we noted as important when Bin looked into this: how much disk space do you think the metadata will require? how much work might resizing support require?
15:45knewt assuming you just want to resize the cow'ed device (and i can't see resizing the backing device making sense), i'd suggest just appending a linear mapping after the cow mapping and then resizing the filesystem
15:46_MarkW Ian had a plan to thread remapping metadata in between CoW'd blocks (!) on the writeable device. That way, some seeks are (hopefully) avoided and the resizing of the writeable volume should Just Work.
15:46_MarkW Do you have problems with the size of your bitmap if resizing is done?
15:47Surferdude getting a file not found on the vmlinuz when bootinf
15:47Surferdude booting
15:48knewt that 2GB example from earlier, with the default chunk size, takes 64Kb of bitmap space on disk. when header code is in i would imagine only 512bytes or 1Kb for the header.
15:48--- ---> muli [] has joined #xen
15:48_MarkW Surferdude: double check your grub options for typos (maybe vmlinux vs vmlinuz?)
15:49Surferdude your right... heh
15:49Surferdude I knew this kvm would come in handy
15:50* knewt finds a terminal server very handy
15:50* _MarkW uses running down the corridor with a hammer instead of a kvm
15:51Surferdude I tried installing Xen on this machine, but it wouldent boot no matter what i tried
15:51riel a hammer? you have to hit that reset button really hard, huh? ;)
15:51_MarkW The server room seems conveniently close until you have to start going there every 2 minutes.
15:51riel nothing like serial console + remote power switch
15:52_MarkW riel: If your machines don't respect you, you got nothin' ;-)
15:52knewt i keep meaning to build my own remote power switch :)
15:52Surferdude lol
15:52_MarkW Surferdude: you must be close to Xenification now though...
15:52Surferdude Yup
15:54_MarkW whoa! just spent an hour on this filesystem code without noticing!
15:54_MarkW time for a typing break, i'll be back in a bit.
15:56Surferdude well... it boots the profile, but still getting the nfs error
15:56Surferdude "unable to mount root fs via nfs"
15:57Surferdude anyone?
15:59knewt what have you used as root= ?
15:59Surferdude root=LABEL=/
15:59knewt that won't work without an initrd
15:59Surferdude ah
15:59knewt just use the actual device name. much easier imho
16:00yosh is initrd known broken with xen 2.0.5?
16:02knewt not that i'm aware. i'm using an initrd a -testing not that much further along than 2.0.5 iirc
16:02knewt s/initrd a/initrd with a/
16:02yosh hm, I was running into trouble
16:03knewt i need it to boot off an evms device :)
16:03yosh Initial guest OS requires too much space
16:03yosh (3500MB is greater than 512MB limit)
16:03yosh ERROR: Error constructing guest OS
16:04Surferdude trying /dev/sda
16:05knewt err. didn't grub-install say it was /dev/hda ?
16:06yosh the initrd is only 4 MB, so I'm not sure where it's getting the 3500MB from
16:12--- <<-- DEac- [] has quit (Ping timeout: 480 seconds)
16:14Surferdude dev/hda isnt working either :(
16:14Surferdude Unable to mount root fs on unknown block
16:15Surferdude help?
16:17_MarkW Surferdude: you need to create some device nodes or a Xen initrd
16:17Surferdude just define another module?
16:17_MarkW yosh: what version of Xen are you running? I remember a bug with initrds a while ago
16:17Surferdude in grub.conf?
16:18_MarkW Surferdude: yes, but you should strictly run mkinitrd to make a Xenified initrd (your exsiting RHEL initrd might work, tho)
16:18yosh _MarkW: 2.0.5
16:19_MarkW yosh: weird. it should work. can you have a quick search of the mail archives for the "requires too much space" or similar errors?
16:20_MarkW i think the problem was resolved there (it might even be fixed in -testing, I guess)
16:20Surferdude run intrd from a shell prompt?
16:20Surferdude mkintrd?
16:21yosh _MarkW: I didn't find anything
16:21_MarkW yosh: definitely there somewhere. it could be in either xen-users or xen-devel, or on the xen-devel list
16:22_MarkW Surferdude: mkinitrd auto-builds an initrd.
16:22_MarkW try the RHEL initrd first tho - might work.
16:26--- ---> DEac- [] has joined #xen
16:26yosh _MarkW: ah, it's on (bleh, I hate the list interface there)
16:27_MarkW yosh: so do I. GMane, I believe, has a unified archive for both the old and new xen-devel lists.
16:28_MarkW yosh: does it give any clues about your problem?
16:28Surferdude This is getting really really annoying
16:28yosh _MarkW: still looking...
16:31_MarkW Surferdude: try mkinitrd /boot/initrd-2.6.11-xen0.img 2.6.11-xen0
16:35Surferdude at a shell prompt
16:35Surferdude ?
16:36_MarkW Surferdude: yup. that'll build you a Xenified RHEL initrd (you might have to tweak the naming depending on the version you've installed)
16:36_MarkW you can then stick that in another module line and it might even work!
16:45knewt _MarkW: ok, i've got some completely unscientific speed test results :) writing out 1gig to a linear mapping took 28seconds. the same with a brand new cow mapping took 35seconds. then filled up the cow mapping filesystem to the max, removed the dummy file, and wrote out the 1gig again. 28.7seconds
16:47_MarkW knewt: sounds promising!
16:51knewt i could probably get it faster by providing an option to not write out the bitmap more frequently then every N jiffies, but i'm not sure if i'd want to create an unsafe option like that
16:51--- <<-- jonmason [] has quit (Read error: Connection reset by peer)
16:53knewt once i've got everything i've already planned up and working, i might look at creating an interwoven bitmap/cow mode
16:54knewt see how much if any speed improvement that might give
16:55_MarkW cool. a decent cow implementation would be very useful to a lot of people
16:58--- ---> jonmason [] has joined #xen
16:59knewt want to be a guineapig^Wvolunteer sometime next week?
17:02_MarkW knewt: i'm interested to see what you come up with but can't guarantee when i'll get a chance to try it out
17:04_MarkW if you send me a link, i can have a poke at some point
17:04_MarkW alternatively, if you post to xen-devel I imagine you will get plenty of feedback ;-)
17:05knewt heh. yeah. i want to do some clean-up at minimum before i give it out for testing, which is why sometime next week
17:05knewt i might even get the background cow-ifying mode in there in that time
17:06_MarkW background cowifying?
17:07_MarkW sounds cool whatever it is...
17:07--- <<-- nextime [] has quit (Quit: BitchX-1.0c19 -- just do it.)
17:09knewt the standard mode is "ondemand", which only copies across from the backing device to the cow device when a write is carried out. "background" mode additionally makes copies of chunks spontaneously from a very low priority thread running in the background
17:10--- <<-- edsuom [] has quit (Quit: Real work beckons)
17:13knewt so you could, for instance, create a new cow mapping for a domain, and use background mode. even if no writes are made to the mapping, eventually the entire contents will be cow-ified across and will increase the event count of the device. so you could have a user task waiting on that event, and when it triggers load a new linear mapping into the inactive table slot and then switch over to it
17:16knewt did that get through?
17:17_MarkW yup
17:18_MarkW i was distracted
17:18_MarkW makes sense.
17:18_MarkW basically a fork() for disks that eventually makes them independent - cool.
17:18knewt cool. my tunnel was down for a couple of minutes due to the annoying 24h bounce. i've never sent text through during that time before
17:19_MarkW 24h bounce?
17:19knewt over here the isp's are gits and even on adsl they like, all, bounce you every 24 hours
17:20_MarkW Ah, that's pretty nasty!
17:20knewt you can't even get a single static IP without paying for a business line
17:20knewt i used to have a /29 on a residential when i was still in the UK
17:24_MarkW my connectivity is via the university, so it's fairly decent
17:26knewt unfortunately i run out of money in about a month or two for the virtual server i pay for, at which point i'll lose the tunnel :(
17:27_MarkW :-(
17:32--- ---> aliguori [] has joined #xen
17:32--- <<-- aliguori [] has quit (Remote host closed the connection)
17:36riel knewt: get a free ipv6 tunnel for your irc connections ;)
17:38_MarkW Surferdude: did you have any luck with mkinitrd?
17:40knewt XenSource don't have any jobs going do they? *g*
17:41riel you might want to ask
17:44--- ---> aliguori [] has joined #xen
17:44--- <<-- aliguori [] has quit (Quit: )
17:44_MarkW knewt: no idea but it might actually be worth asking. they don't seem to advertise as such for tech positions, so if you don't ask, you won't get.
17:45--- <<-- Arnald [] has quit (Ping timeout: 480 seconds)
17:47--- ---> aliguori [] has joined #xen
17:47--- <<-- aliguori [] has quit (Remote host closed the connection)
18:02--- ---> hollis [] has joined #xen
18:36Surferdude back to work...
18:39_MarkW Surferdude: heh :-) let me know how you get on
18:39Surferdude what was that command again?
18:40_MarkW mkinitrd /boot/initrd-2.6.11-xen0.img 2.6.11-xen0
18:40_MarkW or something like that
18:43Surferdude lib/modules/2.6.11-xen0 is not a directory
18:47--- <<-- wenchien [] has quit (Ping timeout: 480 seconds)
18:47_MarkW Surferdude: what is your vmlinuz called?
18:47Surferdude hold on... booting
18:47Surferdude getting a serial input error
18:47Surferdude on booting xen
18:47_MarkW the version number to mkinitrd should be the same as that version
18:47Surferdude diddnt let me do mkinitrd
18:48_MarkW it should work.
18:49Surferdude is there any way to access grub.conf from grub itself
18:49hollis doesn't GRUB have a "cat" command?
18:49Surferdude no clue
18:49hollis well, are you in GRUB now?
18:50knewt yes, it has a cat command
18:51Surferdude title XenLinux 2.0.5
18:51Surferdude root (hd0,0)
18:51Surferdude kernel /xen.gz dom0_mem=131072
18:51Surferdude module /vmlinuz-2.4-xen0 root=/dev/hda ro
18:51Surferdude initrd /intrd-2.4.21-27.el.img
18:52Surferdude Whenever i boot that configuration, i get a serial input error
18:54_MarkW Surferdude: don't use an "initrd" line, use a second module line.
18:54_MarkW the grub initrd line will break things - don't ask me why :-)
18:54--- <<-- soffi_ [] has quit (Ping timeout: 480 seconds)
18:55--- ---> soffi_ [] has joined #xen
18:57--- <<-- JViz [] has quit (Ping timeout: 480 seconds)
19:09Surferdude still getting this stupid serial input error... argh!
19:11Surferdude has anyone experienced this?
19:16_MarkW Surferdude: nope, never heard of it before
19:20Surferdude Ok... changed the kernel back to the way it was and now getting the "cannot mount root fs" message
19:20Surferdude mkinitrd isnt working
19:22knewt i notice it's a 2.4 kernel. is the filesystem designed for devfs and you don't have devfs compiled into the dom0 kernel?
19:23Surferdude how would i find that out
19:24_MarkW yes, i was worried about the 2.4-ness as well
19:24_MarkW Surferdude: what release of RHEL is this?
19:25_MarkW Surferdude: to find out if you're using devfs, type "mount" under an standard boot of RHEL and see if devfs is listed
19:25Surferdude ok
19:25Surferdude rebooting now
19:25Surferdude was trying to see if xenU was working
19:25--- ---> JViz [] has joined #xen
19:26--- <<-- hollis [] has quit (Quit: leaving)
19:27Surferdude [root@plain root]# mount
19:27Surferdude /dev/hda3 on / type ext3 (rw)
19:27Surferdude none on /proc type proc (rw)
19:27Surferdude none on /dev/pts type devpts (rw,gid=5,mode=620)
19:27Surferdude /dev/hda1 on /boot type ext3 (rw)
19:27Surferdude none on /dev/shm type tmpfs (rw)
19:27Surferdude [root@plain root]#
19:28knewt ok. so you should be using root=/dev/hda3
19:30_MarkW Surferdude: this is all very interesting because your problem clearly isn't what I expected :-)
19:30_MarkW where did you get your XenLinux kernel from?
19:30Surferdude i got the prebuilt tarball from
19:31_MarkW ok, use the 2.4-xen0 kernel for now. just set your root=/dev/hda3 in the grub config like knewt said.
19:32Surferdude Ok
19:32Surferdude title XenLinux 2.0.5
19:32Surferdude root(hd0,0)
19:32Surferdude kernel /xen.gz dom0_mem=131072
19:32Surferdude module /vmlinuz-2.4-xen0 root=/dev/hda3 ro
19:33Surferdude I hope that works
19:33_MarkW Surferdude: looks good.
19:34Surferdude back to the serial input error
19:35Surferdude I think that has somthing to do with the kvm
19:36_MarkW do you actually have a serial line attached?
19:36knewt how far is it getting? is it erroring in grub? in xen? in linux?
19:36Surferdude yes its a serial kvm
19:36Surferdude its erroring a few seconds after i select it in grub
19:36_MarkW in Xen itself?
19:37Surferdude I am assuming so
19:37_MarkW what does the error say?
19:37Surferdude Somthing like "serial input error: press ctrl+a three times to enable serial access
19:37Surferdude tried doing that
19:37Surferdude just reboots
19:40Surferdude This is really stumping me :/
19:42knewt what processor do you have?
19:42Surferdude Intel Celeron
19:43_MarkW the ctrl+a thing is normally printed out when Xen starts
19:43--- ---> tierra [] has joined #xen
19:43Surferdude yeah but it just reboots right after it shows
19:43_MarkW are you sure it's a serial error and not just the serial message interleaved with the real error?
19:44Surferdude I cant tell...
19:44knewt the normal message about ctrl-a is: "*** Serial input -> DOM0 (type 'CTRL-a' three times to switch input to Xen)."
19:44Surferdude thats
19:44Surferdude thats it
19:44_MarkW Surferdude: Yup that's normal.
19:44_MarkW So it's not the serial line that's the problem.
19:44knewt Surferdude: the lines immediately above that, do they mention loading domain 0?
19:45_MarkW Some bit of configuration must be out of place.
19:45Surferdude So what is O_o
19:45_MarkW If you put "noreboot" onto the Xen grub command line then it won't reboot automatically (you'll have to power cycle the box)
19:45_MarkW That way you can have a better look at the error.
19:47Surferdude thats when it boots right?
19:47Surferdude when i am @ the grub menu
19:47_MarkW you can press "e" on the Xen menu entry, then "e" to edit the "kernel xen.gz" line, then add it on.
19:47knewt yeah. needs to go on the "kernel" line
19:48_MarkW then press return, then b to boot.
19:51Surferdude Ok... just freezes at the serial input message
19:51_MarkW OK, that's weird. No visible error messages?
19:52demon the linux kernel needs to be told where to direct its console output, it'll want to output it to the local console, not serial...
19:52_MarkW demon: good point! :-)
19:52demon try appending 'console=ttyS0,[speed]' to the kernel command line
19:52demon also, do you have a line in inittab to spawn a getty on the serial port?
19:52demon MarkW: I've installed xen by wire on a few boxes :)
19:53Surferdude no other errors... adding the kernel arg now
19:53demon er, make sure you append that to the linux kernel's command line
19:53_MarkW Surferdude: stick "console=ttyS0" to the end of the vmlinuz line
19:53demon not the xen kernel's
19:53_MarkW works on my test box here
19:54_MarkW you needn't worry about the speed for that line, since Linux doesn't get to control it anyhow.
19:54--- ---> rusty [] has joined #xen
19:55Surferdude same error...
19:55demon no further output?
19:55demon did you build the XenoLinux kernel yourself, or is it some kind of packaged kernel?
19:55_MarkW Surferdude: can you show us what your menu.lst entry currently looks linke?
19:56Surferdude Sure... give me a sec
19:57--- ---> hollis [] has joined #xen
19:58--- <<-- tierra [] has quit (Remote host closed the connection)
19:58knewt heh. looking over my menu.lst file reminds me of when i upgraded my dom0 and suddenly found that the initrd support had been fixed, and that my workaround to deal with the prior bug didn't work any more :/
19:58Surferdude What the hell... i cant even access ssh anymore
20:00Surferdude Crap... might as well get another 24 hours on the kvm
20:01* riel hurts his head on more VM stuff
20:02Surferdude there isnt anything on menu.lst about xen
20:02Surferdude wait... nevermind
20:02Surferdude there is
20:03_MarkW riel: what you up to in the VM? the VFS is hurting my head right now...
20:03Surferdude heres the entry in menu.lst
20:03Surferdude title XenLinux 2.0.5
20:03Surferdude root(hd0,0)
20:03Surferdude kernel /xen.gz dom0_mem=131072
20:03Surferdude module /vmlinuz-2.4-xen0 root=/dev/hda3 ro
20:04_MarkW can you please stick "console=tty0 console=ttyS0" onto the end of that module line?
20:05Surferdude in menu.lst?
20:05_MarkW yeah, it'll need to be there so might as well add it permanently.
20:05demon Surferdude: did you build the XenoLinux kernel yourself, or is this some kind of packaged kernel you're using?
20:06Surferdude its prebuilt
20:06Surferdude title XenLinux 2.0.5
20:06Surferdude root(hd0,0)
20:06Surferdude kernel /xen.gz dom0_mem=131072
20:06Surferdude module /vmlinuz-2.4-xen0 root=/dev/hda3 ro console=tty0
20:06Surferdude console-ttyS0
20:06demon okay... are you sure it supports your IDE controller and NIC?
20:06demon that should be console=ttyS0
20:06demon not console-ttyS0
20:07demon is there an initrd that came with that kernel, perhaps?
20:07_MarkW demon: if we can at least get some output on the serial line then we can figure out what it doesn't like.
20:07_MarkW i don't think it's using udev or anything like that so it shouldn't need an initrd.
20:07Surferdude ok... rebootinng
20:08_MarkW It's probably lacking the IDE controller driver, I guess.
20:09riel _MarkW:
20:09Surferdude its...working!!!!
20:09riel _MarkW: I'm designing a Clock-Pro variation that can be implemented in the Linux VM
20:09_MarkW Surferdude: wicked!
20:10_MarkW we can all give ourselves a pat on the back :-)
20:10demon Surferdude: okay, now is it giving you any kind of error?
20:10_MarkW riel: arrrghhh you scary man :-)
20:10Surferdude Well... it gave me the warning aboout lib/tls
20:10Surferdude but the terminal still shows RHEL
20:10riel _MarkW: want me to forward you the patch I sent out friday ? ;)
20:10demon ok... that would seem to mean it's mounting the filsystem and spawning init
20:10Surferdude at the login prompt
20:10_MarkW Surferdude: OK, that's standard, just move /lib/tls out of the way so it won't get used.
20:11riel if you run Fedora, you want to keep /lib/tls
20:11_MarkW riel: I'd quite like to see the patch actually actually
20:11Surferdude I know that, but should i worry about it still saying RHEL?
20:11riel otherwise RPM database locking doesn't go right
20:11--- ---> nextime [] has joined #xen
20:11Surferdude at the login prompt
20:11Surferdude ?
20:11demon well, you're still running RHEL
20:11riel Surferdude: don't worry about it saying RHEL
20:11demon so yeah, that's normal
20:11riel _MarkW: what's your email address again ?
20:11_MarkW Surferdude: log in and play!
20:11Surferdude ok... good
20:11_MarkW riel:
20:12Surferdude wow... i am so excited that i forgot my root password for a second
20:12_MarkW Surferdude: you should be able to bring up some domains now :-) try the ttylinux example in the user manual for starters - it's nice and simple.
20:13Surferdude xend start brings up a whole lot of errors
20:13Surferdude :/
20:13_MarkW have you installed twisted?
20:13Surferdude yes
20:13Surferdude all python errors
20:13riel Surferdude: make sure to leave /lib/tls in place, if you value /var/lib/rpm/ ;)
20:14knewt eh. i didn't think the 2.4 kernel supported /lib/tls anyway?
20:14_MarkW Surferdude: OK, lets see the Python spew. It's probably something straightforward.
20:14riel knewt: RHEL3 does
20:14knewt aah
20:14_MarkW knewt: they don't but this is a RedHatMegaPatch 2.4 :-)
20:15demon probably just version warnings
20:15riel _MarkW: the reason of that patch (in RHL8 & 9) was to prepare userland for 2.6
20:15Surferdude import error: no module named logging
20:15_MarkW except - it's running a Xen 2.4 kernel now, so I'm not sure why it would be using TLS...
20:15Surferdude thats what it ends with
20:15riel _MarkW: it worked well enough, and there was enough customer demand, that it got included in RHEL3
20:15_MarkW Surferdude: ah, you'll be missing the python logging framework then
20:15riel _MarkW: good point
20:15_MarkW spooky...
20:15Surferdude how do i get that working?
20:16* _MarkW scratches his head
20:16_MarkW what python version are you running?
20:18Surferdude Wont even let me see that :/
20:18_MarkW ?
20:18_MarkW python --version
20:18_MarkW oops, no, that's not it...
20:19_MarkW python -V should tell you
20:19Surferdude 2.2.3
20:20_MarkW Surferdude: you'll probably need to install the logging package as an add-on
20:20_MarkW it wasn't standard until 2.3
20:20knewt ok, i think it's time for me to grab some shuteye, a) it being gone 3am, and b) /me having got a good idea of how i'm going to code interleaving why i come to implement it
20:21Surferdude Where/how do i do that?
20:21_MarkW knewt: see ya.
20:21_MarkW Surferdude: i'm working on it...
20:21* knewt & # to see the wizard...
20:22_MarkW Surferdude: wget
20:22_MarkW tar xzf logging-
20:22_MarkW cd logging-
20:22_MarkW python install
20:24Surferdude host not found
20:24_MarkW ifconfig
20:24_MarkW see if eth0 appears
20:24Surferdude nope
20:25Surferdude ifconfig eth0?
20:25_MarkW you're missing the driver for your network card
20:25_MarkW you'll need to build a new kernel with support for your network
20:25_MarkW if you don't want to do that just yet then you could just reboot into vanilla linux to do the download
20:28_MarkW riel: is anybody actively working on the compressed caching project?
20:28_MarkW i always thought it looked quite cool but couldn't see any obvious signs of life on the website.
20:30Surferdude Looks like I have to go on a download hunt
20:30riel _MarkW: not lately afaik
20:30riel _MarkW: and yes, compressed caching looks very promising
20:31_MarkW does any OS use it presently?
20:31riel not afaik
20:32_MarkW *yawn*
20:32_MarkW Time for me to go to bed!
20:32_MarkW Surferdude: good luck with setting things up. I'll probably be back around the same time tomorrow.
20:33Surferdude Later... thanks for your help
20:33_MarkW riel: thanks for the links / code - I'll check it out when I'm more awake ;-)
20:33* _MarkW goes
20:33--- <<-- _MarkW [] has quit (Quit: Connection interrupted by tinfoil hat)
21:47--- ---> grifferz [] has joined #xen
22:59--- <--- VS_ChanLog [] has left #xen (Rotating Logs)
22:59--- ---> VS_ChanLog [] has joined #xen
23:12--- ---> itamarjp [] has joined #xen
23:26--- <<-- hollis [] has quit (Quit: leaving)
23:43--- ---> soffi [] has joined #xen
23:54--- <<-- rusty [] has quit (Quit: Client exiting)
---Logclosed Sun May 01 00:00:35 2005