Back to Home / #xen / 2005 / 06 / Prev Day | Next Day
#xen IRC Logs for 2005-06-22

---Logopened Wed Jun 22 00:00:21 2005
01:57--- <<-- trash [trash@IPv6.databerlin.org] has quit (Ping timeout: 480 seconds)
01:57--- <<-- movement [~moz@82.152.178.121] has quit (Read error: Connection reset by peer)
02:01--- ---> marl_work_ [~marl@82-40-105-78.cable.ubr01.dumb.blueyonder.co.uk] has joined #xen
02:01--- <<-- marl_work [~marl@82-40-105-78.cable.ubr01.dumb.blueyonder.co.uk] has quit (Read error: Connection reset by peer)
02:02--- ---> trash [trash@IPv6.databerlin.org] has joined #xen
02:08--- <<-- viewbee [~viewbee@198.146.4.74] has quit (Ping timeout: 480 seconds)
02:09--- ---> viewbee [~viewbee@198.146.4.74] has joined #xen
02:10--- ---> movement [~moz@82.152.178.121] has joined #xen
02:18--- ---> Gabo_ [~toby@200.4.252.29] has joined #xen
02:18--- <--- Gabo_ [~toby@200.4.252.29] has left #xen ()
02:25--- <<-- trash [trash@IPv6.databerlin.org] has quit (Ping timeout: 480 seconds)
02:25--- ---> trash [trash@IPv6.databerlin.org] has joined #xen
02:35--- ---> rusty [~rusty@ppp61-60.lns1.cbr1.internode.on.net] has joined #xen
03:22--- User: *** movement is now known as movewent
03:25mael hi rusty_home
03:30rusty mael: hi! Damn internet, everyone knows you're a dog.
03:42--- <<-- rusty [~rusty@ppp61-60.lns1.cbr1.internode.on.net] has quit (Remote host closed the connection)
03:43--- ---> rusty [~rusty@ppp61-60.lns1.cbr1.internode.on.net] has joined #xen
03:48mael good dog, good dog :)
03:49* rusty feels he missed something by pressing Control-Alt-BS at the wrong time.
03:49mael hehe no you didn't
03:50mael well, not on irc at least :)
03:57--- <<-- rusty [~rusty@ppp61-60.lns1.cbr1.internode.on.net] has quit (Quit: Client exiting)
03:59--- <<-- movewent [~moz@82.152.178.121] has quit (Remote host closed the connection)
04:40--- ---> imul [~muli@nesher3.haifa.il.ibm.com] has joined #xen
04:45--- ---> athomas [~athomas@ppp-0-91.lond-b-1.access.uk.tiscali.com] has joined #xen
04:54--- <<-- DEac- [~deac@xdsl-81-173-161-168.netcologne.de] has quit (Ping timeout: 480 seconds)
05:01--- <<-- zimbo [~zimbo@callisto.dom.bonis.de] has quit (Quit: )
05:05--- ---> DEac- [~deac@xdsl-213-196-207-144.netcologne.de] has joined #xen
05:18--- ---> zimbo [~zimbo@callisto.dom.bonis.de] has joined #xen
05:28--- Netsplit uranium.oftc.net <-> keid.oftc.net, arion.oftc.net quits: aeddan, @Sir_Ahzz, hbaum__, thelsdj, knewt, kmacy, surriel, lilo, albeiro, dhendrix, (+2 more, use /NETSPLIT to show all of them)
05:29--- Netsplit over, joins: thelsdj, DEac-, aeddan, Sir_Ahzz, kmacy, knewt
05:29--- Netsplit over, joins: lilo, dhendrix, albeiro, newsham, hbaum__, surriel
05:29--- Channel: ServerMode/#xen [+o Sir_Ahzz] by keid.oftc.net
06:27--- <<-- soffi [~soffi@Tolvudeild-SC-gw.vv.simnet.is] has quit (Ping timeout: 480 seconds)
06:27--- <<-- tessier [~treed@222.253.84.134] has quit (Ping timeout: 480 seconds)
06:32--- ---> soffi [~soffi@Tolvudeild-SC-gw.vv.simnet.is] has joined #xen
07:11--- ---> Sir-Ahzz [~ahzz@c-24-0-232-12.hsd1.tx.comcast.net] has joined #xen
08:02--- ---> deac [~deac@xdsl-213-196-198-138.netcologne.de] has joined #xen
08:05--- <<-- DEac- [~deac@xdsl-213-196-207-144.netcologne.de] has quit (Ping timeout: 480 seconds)
08:13--- <<-- katzj [~katzj@c-66-30-107-17.hsd1.ma.comcast.net] has quit (Ping timeout: 480 seconds)
08:14--- ---> cilkay [~cilkay@CPE00d0b743a22f-CM0011ae01fcbe.cpe.net.cable.rogers.com] has joined #xen
08:15cilkay Hi. When I boot dom0, I see messages about missing ehci modules to support USB. At the moment, that isn't important to me but it might be in the future. Do I have to build a kernel from scratch to build those modules?
08:17tab cilkay: probably
08:17cilkay thanks tab
08:40--- <<-- sdague [~sdague@user-0cevegb.cable.mindspring.com] has quit (Ping timeout: 480 seconds)
08:42--- User: *** unriel is now known as riel
08:51--- ---> katzj [~katzj@wlanconf-nat-pool-bos.redhat.com] has joined #xen
09:01--- ---> hollis [~hollis@user-0vvde2g.cable.mindspring.com] has joined #xen
09:04--- <<-- Sir-Ahzz [~ahzz@c-24-0-232-12.hsd1.tx.comcast.net] has quit (Quit: Client exiting)
09:04--- ---> rharper [~rharper@pixpat.austin.ibm.com] has joined #xen
09:08--- <<-- muli [~muli@alhambra.mulix.org] has quit (Ping timeout: 480 seconds)
09:16--- ---> sdague [~sdague@bi-02pt1.bluebird.ibm.com] has joined #xen
09:38--- <<-- imul [~muli@nesher3.haifa.il.ibm.com] has quit (Quit: BitchX: born to raise hell)
09:50--- ---> aliguori [~anthony@cpe-70-112-81-91.austin.res.rr.com] has joined #xen
10:27--- <<-- aliguori [~anthony@cpe-70-112-81-91.austin.res.rr.com] has quit (Quit: Leaving)
10:35--- ---> mejlholm [~mejlholm@port79.ds1-abc.adsl.cybercity.dk] has joined #xen
10:47cilkay I am confused over what the rootfs should be for domU. I created a partition, /dev/sda9, which I designated as / for domU. I would have used a LVM volume were it not for the fact that the Mandriva installer complains that it cannot put mount LVM volumes as root partitions.
10:47cilkay I have: disk = [ 'phy:/dev/sda9,sda1,w' ]
10:47cilkay root = "/dev/sda9 ro"
10:47cilkay in the config file for domU.
10:48cilkay I guess I should make that /dev/sda1
10:48demon no, it should be '/dev/sda1'
10:48demon because that's the designation within the new domain for the device it sees
10:48demon the 'disk =' line is correct, but you'll need to fix the 'root =' line
10:49cilkay And fstab within domU presumably.
10:49demon correct
10:53cilkay I have: /dev/sda1 / reiserfs notail 1 1 in domU's /etc/fstab
10:53--- ---> aliguori [~anthony@pixpat.austin.ibm.com] has joined #xen
10:53demon er... reiser? I've heard of people having trouble with reiser in domU
10:53cilkay disk = [ 'phy:sda9,sda1,w' ] and root = "/dev/sda1 ro" in the config file in dom0
10:54cilkay ok, I'll try ext3 then
10:54demon I've used ext2, ext3, jfs, and xfs with success though
10:54demon well, have you tried booting it as is yet?
10:55cilkay Yes, I get a kernel panic, VFS: Cannot open root device "sda1" or unknown-block(2,0)
10:55cilkay Please append a correct "root=" boot option
10:55cilkay Kernel panic - not syncing: VFS: Unable to mount root fs on unknown-block(2,0)
10:55cilkay Is there a way to convert the fs without destroying the contents? It's not critical that I preserve domU and can reinstall if necessary so I don't care if it is risky.
10:56demon you don't have devfs built into that domU kernel, right?
10:56trash cilkay: yes, by backing it up and recovering it.
10:56cilkay demon, I don't know. I did a binary install.
10:56cilkay trash, thanks
10:57demon and do you have SCSI disk support built into that kernel? it sounds like it's getting the wrong major/minor number
10:57demon 2/0 is /dev/hda
10:57cilkay I do.
10:57demon 8/1 should be /dev/sda1
10:57cilkay It's a serial ATA drive. It is seen as a SCSI device.
10:57demon well, I'm talking about the domU kernel
10:57demon not the dom0 kernel
10:58cilkay Ah. Don't know. I'm using the binary domU kernel.
10:58demon try just changing references to 'sda1' to 'hda1' in the xen domain definition (and the domU's fstab) and see what happens...
11:00cilkay Nope, same thing.
11:00cilkay This time complaining about hda1
11:00demon does it have the same major/minor numbers?
11:00cilkay 2, 0
11:00cilkay Yes, same.
11:01demon that's... uh, intersting
11:01demon interesting, too
11:01--- ---> muli [~muli@alhambra.mulix.org] has joined #xen
11:02cilkay Maybe it is how I'm creating domU. I created the LVM volumes from dom0, rebooted, and then did a standard install targeting /dev/sda9 as / and the rest of it on the LVM volumes.
11:02cilkay The only other domU I ever got working was a file backed one where I just copied a running dom0.
11:18--- <<-- hollis [~hollis@user-0vvde2g.cable.mindspring.com] has quit (Quit: leaving)
11:20--- <<-- dhendrix [~dhendrix@bi01p1.co.us.ibm.com] has quit (Quit: Leaving)
11:28--- ---> dhendrix [~dhendrix@bi01p1.co.us.ibm.com] has joined #xen
11:28--- ---> woody [~marvel@bi01p1.co.us.ibm.com] has joined #xen
11:31--- <<-- woody [~marvel@bi01p1.co.us.ibm.com] has quit (Quit: )
11:32--- ---> woody [~woody@bi01p1.co.us.ibm.com] has joined #xen
11:35--- <<-- enum [~Administr@ip-207-145-127-226.lax.megapath.net] has quit (Read error: Connection reset by peer)
11:49--- ---> hollis [~hollis@pixpat.austin.ibm.com] has joined #xen
11:54--- ---> malahal [~malahal@bi01p1.co.us.ibm.com] has joined #xen
11:56--- ---> tessier [~treed@222.253.78.207] has joined #xen
11:58--- <<-- rread_ [~robert@datarithm.com] has quit (Ping timeout: 480 seconds)
12:01--- ---> rread [~robert@datarithm.com] has joined #xen
12:12--- <<-- athomas [~athomas@ppp-0-91.lond-b-1.access.uk.tiscali.com] has quit (Quit: Leaving)
12:12--- ---> niv [~nivedita@bi01p1.co.us.ibm.com] has joined #xen
12:15--- ---> jimix [~jimix@ip13.194.susc.suscom.net] has joined #xen
12:49--- <<-- muli [~muli@alhambra.mulix.org] has quit (Ping timeout: 480 seconds)
12:51--- <<-- rread [~robert@datarithm.com] has quit (Ping timeout: 480 seconds)
12:52--- ---> rread [~robert@207.47.36.146] has joined #xen
13:02--- <<-- rread [~robert@207.47.36.146] has quit (Ping timeout: 480 seconds)
13:04--- ---> rread [~robert@207.47.36.146] has joined #xen
13:12--- <<-- rread [~robert@207.47.36.146] has quit (Ping timeout: 480 seconds)
13:23--- ---> rread [~robert@207.47.36.146] has joined #xen
13:29--- ---> malahal_ [~malahal@bi01p1.co.us.ibm.com] has joined #xen
13:29--- <<-- malahal [~malahal@bi01p1.co.us.ibm.com] has quit (Quit: Leaving)
13:30malahal_ I have allocated 2G to dom0 (I have 5G physically on the machine)
13:30--- <--- malahal_ [~malahal@bi01p1.co.us.ibm.com] has left #xen ()
13:31--- ---> malahal [~malahal@bi01p1.co.us.ibm.com] has joined #xen
13:39cilkay Is there a recommended partitioning scheme for dom0 or should I just use whatever I would normally use, i.e. separate /boot, /, /usr, /var, /tmp, /home ?
13:40grifferz you're probably not going to have much running in dom0, so I don't think it really matters to use less partitions
13:42--- ---> rread_ [~robert@datarithm.com] has joined #xen
13:43cilkay grifferz, you're right, I'm doing a minimal installation, which is just under 300M. However, I thought keeping logs in a separate partition in /var in order to isolate those things that change from those things that don't still seems like a good idea.
13:43cilkay Same for /tmp
13:43cilkay I could do without /home
13:43cilkay .. as users will not be saving files in dom0 anyway
13:45--- <<-- rread [~robert@207.47.36.146] has quit (Ping timeout: 480 seconds)
13:48grifferz you may want to separate /var if you think it is possible logging could fill up all available space, but I wouldn't separate /home unless there were significant user files, nor /tmp unless I wanted it mounted with special options (noexec, nosuid, ...)
13:53cilkay grifferz, /tmp is mounted with special options if you run Mandrake/Mandriva at msec 5
13:54cilkay ... which causes problems for some apps, though none that I will be running in dom0
13:54cilkay One that comes to mind is Trac - it didn't like the restrictive permissions on /tmp at msec 5
14:00--- <<-- malahal [~malahal@bi01p1.co.us.ibm.com] has quit (Quit: Leaving)
14:03knewt 300 is rather a lot for a minimal installation
14:04cilkay knewt, Fedora Core weighed in at 800M for a "minimal" install so I'm happy with 300M.
14:08knewt i consider 150 megs a reasonable "without effort" miminal installation
14:09--- <<-- katzj [~katzj@wlanconf-nat-pool-bos.redhat.com] has quit (Remote host closed the connection)
14:09cilkay Using which distro?
14:10cilkay I'm sure I could cut mine down, if I wanted to live without urpmi, which I don't.
14:13knewt that's on ubuntu/warty. if i went for debian/woody (easier if using vmtools instead of xend), then it'd probably be even lower
14:18cilkay knewt, the Sarge installer choked on my hardware. No idea why and I wasn't particularly motivated to fix it.
14:42--- <<-- aliguori [~anthony@pixpat.austin.ibm.com] has quit (Quit: Leaving)
14:44--- ---> aliguori [~anthony@pixpat.austin.ibm.com] has joined #xen
14:49--- ---> muli [~muli@alhambra.mulix.org] has joined #xen
14:52--- ---> aliguori- [~anthony@pixpat.austin.ibm.com] has joined #xen
14:52--- <<-- aliguori [~anthony@pixpat.austin.ibm.com] has quit (Read error: Connection reset by peer)
14:52--- User: *** aliguori- is now known as aliguori
15:03--- ---> malahal [~malahal@bi01p1.co.us.ibm.com] has joined #xen
15:05--- ---> cdub [~chrisw@216-99-217-87.dsl.aracnet.com] has joined #xen
15:12--- <<-- malahal [~malahal@bi01p1.co.us.ibm.com] has quit (Ping timeout: 480 seconds)
15:34--- <<-- muli [~muli@alhambra.mulix.org] has quit (Ping timeout: 480 seconds)
15:38--- ---> malahal [~malahal@bi01p1.co.us.ibm.com] has joined #xen
15:48zut|afk good evening all
15:49zut|afk how do i nicely detech a console instead of kill `pidof console 32` ?
15:49zut|afk detach
15:50rharper if you are using xm console, you can press <ctrl> ']'
15:51zut|afk ahhh...nice...thx:)
15:52--- ---> MarkWilliamson [~MarkW@maw48.kings.cam.ac.uk] has joined #xen
15:52rharper sure
15:52zut|afk just to determine wether i am stupid or not...could i have found that in the docs?
15:52rharper I don't recall, I think I learned that from using telnet
15:52rharper and it just worked on the console
15:53zut|afk ok
15:53cilkay I did a fresh install of Mandriva 2005 LE. During the installation, I created an LVM volume of 5GB. I am now running a Xen kernel in dom0. I subsequently did a fresh installation in the filesystems I had made in the LVM, wha/root, wha/usr, wha/var, and wha/home.
15:53MarkWilliamso| zut|afk: Anyone who patches any holes in the docs has my everlasting gratitude ;-)
15:54zut|afk heheh..i can imagine.... you have enuf work coding right? :)
15:54cilkay When I attempt to start domU, I get: xen_blk: Initialising virtual block device driver unknown partition table
15:55cilkay MarkWilliamson, I'm attempting to do a step by step HOWTO for Mandriva here: http://www.dinamis.com/blog/
15:55MarkWilliamso| zut|afk: yeah! I'm spending all my time banging my head against the keyboard at the moment. Just been tracking down some really obscure bugs.
15:55MarkWilliamso| cilkay: Excellent!
15:55zut|afk hehe
15:55zut|afk carefull with the banging...chances are you need your head desperately solving the bugs
15:55MarkWilliamso| cilkay: How are you exporting devices to the guest? i.e. what's your "disks = " line when it fails?
15:56MarkWilliamso| zut|afk: Ah, that's what it's for! I'd been wondering, since it didn't seem to do much :-)
15:56zut|afk hehehe
15:56MarkWilliamso| I can now almost kexec a guest if I restart Xend (don't ask!)
15:57zut|afk bout the tiny piece of knowledge missing from the docs...it was rharper who enlightenend (not a native speaker can u tell:) ) me on how to detach from a console
15:57aliguori MarkWilliamson: nice!
15:57cilkay isk = [ 'phy:wha/root,wha/root,w' ]
15:57cilkay disk = [ 'phy:wha/root,wha/root,w' ]
15:57cilkay root = "wha/root ro"
15:57demon wha?
15:57demon what's that?
15:58cilkay just the name of the LVM volume group
15:58cilkay It's meaningful to me.
15:58MarkWilliamso| cilkay: you can't have the target device (the guest device name) as wha/root
15:58demon ah... well, that probably won't work too well for the device to export as into the target domain
15:58demon because those will have a very specific major/minor that won't be accessible in the domU
15:58MarkWilliamso| cilkay: try sda1 instead (and correspondingly change the root= entry too)
15:59cilkay ok, will do.
15:59demon sda1 or hda1
15:59zut|afk gotta run...bb2morrow
15:59MarkWilliamso| cilkay: obviously, keep the source device the same or it won't get exported ;-)
15:59MarkWilliamso| zut|afk: cya
16:00cilkay Are the target device names arbitrary? IOW, I can just call them sda1, sda2, sda3, etc.?
16:01demon yep
16:01demon well, the only real requirement is that they map to known major/minor pairs
16:01cilkay It's a safe bet that sda* will do that I guess.
16:01demon yep
16:01MarkWilliamso| aliguori: guest kexec isn't that hard conceptually, it's mostly just fiddly. but it'll work in a day or two.
16:02aliguori MarkWilliamson: how do you notify xend once a kexec has happened?
16:02aliguori or do you not and just never send the driver_status messages until the new guest comes up?
16:03MarkWilliamso| aliguori: I thought about a few ways of doing kexec
16:03MarkWilliamso| but letting a guest kexec "internally", i.e. without assistance just duplicates a load of code from the domain builder
16:03aliguori yup
16:04MarkWilliamso| so right now, guests can kexec by sending a normal control message saying "Hey Xend please kexec me now, the new kernel is here"
16:04aliguori MarkWilliamson: does xend have to read the kernel off of the domU filesystem?
16:04MarkWilliamso| aliguori: nope, directly out of memory using a well-defined data structure.
16:05aliguori awesome
16:05cilkay What about swap? Could it be an LVM volume or should I use an existing physical partition? Do I export just like the other partitions from dom0?
16:05aliguori MarkWilliamson: sounds like it will work well
16:05MarkWilliamso| cilkay: export it as usual. it can be backed by anything you want in dom0.
16:05demon it can be an LVM volume
16:05demon that's what I use, and it works just fine
16:05MarkWilliamso| aliguori: the next job after that is to merge kexec in Xen so that doing it in dom0 will restart the machine with a new kernel
16:05cilkay I presume it is not a good idea to share a swap partition amongst various virtual machines.
16:06demon ... no
16:06demon that would be probably worse than trying to share a r/w filesystem
16:06MarkWilliamso| cilkay: never share any block device between virtual machines unless they'll *all* have read-only access
16:06aliguori MarkWilliamson: dom0 kexec??? that would be fantastic :-)
16:06cilkay ok
16:06demon having individual domains overwriting eachother's pages... ugh
16:06demon that would be ugly
16:06MarkWilliamso| cilkay: also, never mount a filesystem writeably in dom0 after starting domains that are reading it
16:07MarkWilliamso| aliguori: my work has come at a good time -kexec is getting merged in 2.6.13
16:07aliguori MarkWilliamson: have you thought about being able to kexec Xen itself?
16:07aliguori that would be neat
16:07aliguori :-)
16:08demon kexec?
16:08MarkWilliamso| aliguori: that's what'll happen if you run kexec in dom0
16:08demon what will that do?
16:08MarkWilliamso| demon: Linux kexecs a new kernel and jumps straigt into it
16:09aliguori MarkWilliamson: oh? see, I was thinking more along the lines of kexec()'ing Xen but not bringing down the domains.. so you can do a live update of Xen
16:09malahal sure, I will.
16:09MarkWilliamso| demon: on bare hardware, saves reboot time in the bios. in Xen its usefull for a few things.
16:09MarkWilliamso| aliguori: I'd thought of that. I'd like to do it but it'll depend if I can convince anyone it's worthwhile
16:10MarkWilliamso| aliguori: Xen isn't supposed to *have* bugs :-)
16:10newsham what magic software technology achives that, mark?
16:10aliguori MarkWilliamson: of course not :-)
16:10demon er... what's the advantage? I don't imagine you can maintain states of block devices and processes and whatnot
16:11MarkWilliamso| aliguori: it'd need a fair bit of work because Xen would have to serialise a load of state about the running domains, then recover it later. but much more doable than the equivalent in a full blown OS. eventually we could be able to hot replace every component of the system!
16:11newsham is the VMX builder intended for vanderpool only or a general loader?
16:11MarkWilliamso| newsham: magic for what? the kexec?
16:11newsham seems like the builders all share a lot of code :\
16:11cilkay I made the changes to the exported block devices and modified fstab in domU accordingly. When I attempt to start, this time it thinks it should be mounting the root from an NFS Server, which is not what I had specified in the conf file.
16:12MarkWilliamso| demon: it's faster than going through the BIOS. and you can jump into a fresh kernel when the main one panics, then map its memory and do a crashdump.
16:12demon hm... I see
16:12aliguori MarkWilliamson: yeah, that's my thoughts exactly.. it's not easy, but it should be doable
16:12demon cilkay: usually if that happens, it means it couldn't open what you told it to open for a root device, so it's falling back to NFS-root
16:13MarkWilliamso| demon: and it enables things like XenoBoot, which really is the pinnacle of bootloaders, IMO
16:13demon XenoBoot?
16:13demon I haven't heard of this
16:13aliguori lol
16:13aliguori pinnacle of bootloaders :-)
16:13MarkWilliamso| XenoBoot - A RedHat 9 liveCD is your bootloader.
16:13demon ...
16:13MarkWilliamso| Usually just kexecs the OS you wanted to load, like a normal bootloader.
16:14MarkWilliamso| But in the case of failures, you can ssh to XenoBoot itself (because it's just Linux) and fix / diagnose things :-)
16:14MarkWilliamso| Much more fun than fooling around on a grub prompt!
16:14cilkay The kernel for domU is installed in /boot of dom0. I am not exporting /boot, read only or otherwise. Should I mount /boot read only in dom0 and export it as read only to domU?
16:14demon no
16:15MarkWilliamso| cilkay: no, the domain builder will poke the kernel into memory itself
16:15demon heh
16:15demon that would be a serious chicken/egg problem
16:15demon nothing running to load the kernel image into the new domain's memory area :)
16:15cilkay I remember this from the ltsp project, pivot_root and such.
16:15aliguori this is the sort of stuff that higher-end boxes that do virtualization already have
16:17MarkWilliamso| I've not actually used XenoBoot myself but it's on all the University's test boxes. I think it's also used in planetlab sometimes but I'm not sure where.
16:17cilkay Aha. From /var/log/xend-debug.log: VmError: Kernel image does not exist: /boot/vmlinuz-2.6.10-xenU
16:18MarkWilliamso| cilkay: does it? are you sure it's not 2.6.11?
16:18cilkay It is 2.6.11 but I thought I had changed it.
16:19cilkay kernel = "/boot/vmlinuz-2.6-xenU" is what I have in the conf file
16:19cilkay ll /boot/vmlinuz-2.6-xenU
16:19cilkay lrwxrwxrwx 1 94 141 22 Jun 23 20:28 /boot/vmlinuz-2.6-xenU -> vmlinuz-2.6.11.10-xenU
16:19cilkay That should work.
16:23MarkWilliamso| cilkay: it sounds like there's a subtle typo somewhere
16:23cilkay Changed it to reference the kernel image directly, same thing.
16:24cilkay I see this in the debug log: xfrd: error while loading shared libraries: libcurl.so.2
16:24MarkWilliamso| Does the error message contain the right path?
16:24demon heh
16:24MarkWilliamso| cilkay: well you do need that library - so maybe that's it!
16:24demon that'd be a slight problem
16:24MarkWilliamso| maybe that's the file that's not being found :-)
16:24demon running debian or an RPM derivative?
16:24cilkay Mandriva
16:25demon mmm... an RPM derivative
16:25cilkay urpmf libcurl.so.2 reports nothing
16:25demon www.rpmfind.net?
16:25cilkay urpmf is the recommended way
16:26cilkay urpmf libcurl.so ==> libcurl3:/usr/lib/libcurl.so.3
16:26MarkWilliamso| cilkay: you did a binary install?
16:26cilkay Yes.
16:26demon ... maybe just make a symlink to /usr/lib/libcurl.so.3 named /usr/lib/libcurl.so.2, and see if that makes it happy
16:26cilkay install.sh was seriously broken btw.
16:26demon heh
16:26MarkWilliamso| I guess this suggests we need some mandriva packages!
16:26demon I always build from source myself
16:27cilkay I had to create a bunch of directories manually.
16:27MarkWilliamso| cilkay: that's due to a weird semantic change in Python at some point
16:27MarkWilliamso| it stopped creating directories automatically. weird!
16:30MarkWilliamso| demon: I build Xen from source about a hundred times a day. It gets old after a while :-)
16:31cilkay Some progress: VFS: Cannot open root device "sda1" or unknown-block(2,0)
16:31cilkay Please append a correct "root=" boot option
16:31cilkay And of course, a kernel panic
16:32demon hm... can you try specifying 0801 as the root device, instead of /dev/sda1? maybe it'll get your drift if you specify it that way
16:32cilkay Major, minor numbers don't match /dev/sda1, which is 8,1
16:33cilkay Sure.
16:33cilkay I'm sure this is because of the way that I installed.
16:33--- ---> chrish01 [~chrish01@69.90.131.10] has joined #xen
16:33--- <<-- chrish01 [~chrish01@69.90.131.10] has quit (Quit: )
16:33cilkay I just did a normal install into the LVM.
16:33cilkay ... but didn't install the bootloader of course.
16:33--- ---> chrish01 [~chrish01@69.90.131.10] has joined #xen
16:34MarkWilliamso| cilkay: does mandriva use udev?
16:34cilkay yes
16:34MarkWilliamso| i bet it does.
16:34MarkWilliamso| right. you'll need to create a few device nodes in the guest's filesystem or it will get confused and not boot.
16:34--- <<-- aliguori [~anthony@pixpat.austin.ibm.com] has quit (Read error: Connection reset by peer)
16:35--- ---> aliguori [~anthony@pixpat.austin.ibm.com] has joined #xen
16:35MarkWilliamso| http://wiki.xensource.com/xenwiki/XenFaq
16:35cilkay Same thing after changing to 0801 except it now complains about 8.1
16:35cilkay ok, will look
16:35demon well, the messages it's giving don't even indicate that it's getting that far...
16:36MarkWilliamso| Actually I agree with demon. probably ignore what I just said.
16:36MarkWilliamso| ummmmm. I know I've seen this before but I can't remember where.
16:37MarkWilliamso| cilkay: can you search the mailing archives for unknown-block(2,0) - that might reveal something.
16:37cilkay k
16:38demon did the prebuilt tarballs ever stop coming with domU kernels with devfs support built in? that was one problem I kept banging into early in 2.0, when I'd try them...
16:38--- <<-- rharper [~rharper@pixpat.austin.ibm.com] has quit (Quit: Leaving)
16:38MarkWilliamso| demon: that's actually what it sounds like to me
16:38aliguori oh, unknown-block.. sounds root= is not right
16:39MarkWilliamso| demon: but I can't understand why it ever was in there!
16:39MarkWilliamso| cilkay: does the boot output mention anything about devfs?
16:39aliguori gentoo really doesn't like not having devfs :-(
16:39demon heh
16:39demon <3 debian
16:39MarkWilliamso| aliguori: it works ok with udev, doesn't it?
16:39cilkay No, I haven't seen that.
16:39demon MarkW: I never could figure that one out
16:39jonmason aliguori: I used udev
16:40cilkay I have a domU working on another Mandriva box, well, sort of.
16:40demon all I know is when I started building the dom0 and domU kernels myself, everything started working right
16:40aliguori MarkWilliamson: not in my config--using LVM for /
16:40jonmason MarkWilliamson: yes, it does
16:40demon so I've just done that ever since
16:40MarkWilliamso| aliguori: ah
16:40cilkay I installed it by just copying dom0 into a file backed VM.
16:40aliguori jonmason: the genkernel initrd assumes devfs...
16:40cilkay err, vbd
16:40jonmason aliguori: GENKERNEL!!!!!!!!!!!!!!!!!!!!!!!
16:40jonmason I hate genkernel
16:41jonmason worst thing to happen to gentoo
16:41jonmason everyone has problems
16:41jonmason compile your own kernel, and everything is hunkydory
16:42demon hah
16:42demon I'd use centos before I'd use gentoo... but I guess that's just me
16:43jonmason demon: don't hate
16:44newsham "dont hate"? what is this, a UPN sitcom?
16:44--- User: *** riel is now known as unriel
16:44demon <- h8r
16:47cilkay Nothing useful in the archives for that error. Most deal with getting dom0 running. I noticed something in /dev of domU.
16:48cilkay Device (2, 0) is /dev/fd0
16:48aliguori what's your root parameter?
16:49cilkay root = "0801 ro"
16:49cilkay I had sda1 in place of 0801 before.
16:50--- <<-- hollis [~hollis@pixpat.austin.ibm.com] has quit (Quit: leaving)
16:58demon yeah... my money is on devfs problems
17:00cilkay So how do I work around that?
17:01cilkay devfs is not unique to Mandriva so someone must have experienced this before - let me check the list archives for devfs
17:01aliguori cilkay: just enable CONFIG_DEVFS in the config
17:01aliguori I'm using it :-)
17:02aliguori it works quite happily
17:02cilkay Mandriva or devfs?
17:02aliguori devfs on gentoo
17:03MarkWilliamso| devfs works ok in dom0 but IIRC it breaks domU
17:04cilkay grep -R DEVFS . in /etc/xen yielded nothing
17:04sdague cilkay: Mandriva is now udev
17:05sdague i've got mandriva 10.2 (udev) running with Xen on a box. File backed vbd
17:05MarkWilliamso| aliguori: aha! All the kexec machinery now appears to work, except for the need to restart Xend :-)
17:06cilkay sdague, so do I. I'm now trying to do an LVM backed vbd
17:06sdague cilkay: if you want to build a fresh domU root, check out http://wiki.sisuite.org/BuildingLinuxChroots
17:07cilkay sdague, building one as opposed to doing a regular install?
17:07sdague cilkay: ah, ok. My total experience with Linux LVM has been not great, and I never dug quite far enough to figure out why
17:07sdague cilkay: you can build a chroot without needing to bring up the installer
17:07sdague just use urpmi
17:08cilkay sdague, I've done it. It is not a trivial exercise as there are some RPMs which don't install properly, though most wouldn't matter for domU.
17:08sdague hmmm, I didn't see that in 10.2
17:09sdague anyway, good luck. I'm off
17:09cilkay sdague, I did a remote install of 10.1 on a colocated box with no remote console or remote hands. It was tricky.
17:15aliguori MarkWilliamson: awesome!
17:17--- <<-- sdague [~sdague@bi-02pt1.bluebird.ibm.com] has quit (Ping timeout: 480 seconds)
17:18--- <<-- mejlholm [~mejlholm@port79.ds1-abc.adsl.cybercity.dk] has quit (Quit: Client exiting)
17:27--- ---> muli [~muli@alhambra.mulix.org] has joined #xen
17:44cilkay Some progress: I did an installation to a physical partition, not an LVM volume. I was able to start domU, though I get the expected errors about missing modules and multiple unexpected "INIT: Id "2" respawning too fast: disabled for 5 minutes" messages to the console.
17:47MarkWilliamso| cilkay: sounds like good news.
17:57--- <<-- chrish01 [~chrish01@69.90.131.10] has quit (Quit: Leaving)
18:01--- <<-- aliguori [~anthony@pixpat.austin.ibm.com] has quit (Ping timeout: 480 seconds)
18:06cilkay MarkWilliamson, I have a feeling that Mandriva does not like to have / on an LVM volume. I'm going to try it on a physical volume with the rest of it on LVM and see if that makes a diff.
18:11--- <<-- niv [~nivedita@bi01p1.co.us.ibm.com] has quit (Quit: Quitting)
18:16cdub cilkay: you'll need an initrd for lvm
18:16MarkWilliamso| cilkay: It doesn't matter if the VBD is on an LVM volume because it'll be transparent tno the guest
18:16MarkWilliamso| cilkay: but (cdub just beat me to it) if the domain *expects* to boot off LVM it'd need an initrd
18:17MarkWilliamso| It's probably preferable for management purposes to have the underlying LVM-ness hidden from the guest though.
18:17cdub yeah
18:18MarkWilliamso| cdub: ah, you're Chris Wright. Hello! *waves*
18:18cdub MarkWilliamson: hi Mark ;-)
18:19MarkWilliamso| cdub: how's things? what are you working on at the moment?
18:19cdub MarkWilliamson: i'm with you (in spirit)...beating head against console or keyboard chasing subtleties ;-)
18:19cdub MarkWilliamson: doing x86_64 smp ATM
18:20MarkWilliamso| cdub: ah - bet that's um... "fun" :-)
18:20cdub MarkWilliamson: well, just finished releasing 2.6.12.1, so now, back to x86_64 smp stuf
18:20cdub MarkWilliamson: hehe, indeed
18:20cdub MarkWilliamson: i spent hours staring at one fscking line of assembly
18:21cdub i think it's permanently burned into my brain ;-)
18:21MarkWilliamso| cdub: I spent most of today chasing a heap corruption bug in the kexec tools
18:21cdub ugh!
18:21MarkWilliamso| I'll have to send them a patch at some point...
18:21cdub MarkWilliamson: heh, or a letter bomb ;-))
18:22MarkWilliamso| It was only a tiny thing but it mean the tools kept segfaulting for no apparent reason.
18:22cdub MarkWilliamson: is that for booting domains in general?
18:22cdub yeah, that's annoying!
18:22MarkWilliamso| I reckon it didn't affect x86(_64) kexec because of a quirk of the memory allocator
18:23MarkWilliamso| It'd probably bite on IA64 because they use a less complex memory map in kexec. :-( it's soooooooo unpleasant!
18:23cdub sounds unpleasant
18:24--- ---> chrish01 [~chrish01@69.90.131.10] has joined #xen
18:24MarkWilliamso| Anyhow, guest kexec should work sometime in the next day or so. Whole host kexec some time later on.
18:24cdub MarkWilliamson: and that'll be preferred way for booting?
18:24MarkWilliamso| cdub: it'll be an alternative to using pygrub as a kernel-chooser
18:25cdub MarkWilliamson: ah, gotcha
18:25cdub MarkWilliamson: i recall it being mentioned in cambridge, but forgot why ;-)
18:26MarkWilliamso| cdub: I'll probably write a little "grub-like" at some stage. It'll also let you run XenoBoot in a domain, which might be cool.
18:26--- ---> aliguori [~anthony@cpe-70-112-81-91.austin.res.rr.com] has joined #xen
18:27cdub nice
18:29cdub be good if kexec gets merged upstream
18:30newsham MW: isnt that what pygrub is supposed to be for?
18:31MarkWilliamso| cdub: Andrew Morton has been making positive noises regarding merging it in 2.6.13 - which would be pretty good timing from our PoV :-)
18:31MarkWilliamso| newsham: yes.
18:31cdub MarkWilliamson: that reminds me, i meant to chime in...
18:33surriel there seems to be around 3-4 months between 2.6.N releases
18:33surriel so I guess it would be good to try to get xenolinux in before 2.6.13 ;)
18:33cdub surriel: yeah
18:38cilkay Progress on the LVM backed VM with a physical device for /: If I comment out the exports for everything but the root device, the VM will boot, of course with a bunch of errors as it cannot find /usr, /var, etc.
18:40cilkay How do you specify multiple vbds to be exported in the conf file?
18:40cilkay I have one line per vbd.
18:41surriel disk = [ 'phy:vg0/rawhide32-guest,sda1,w',
18:41surriel 'phy:vg0/rawhide32-guest-swap,sda2,w' ]
18:41surriel like that ;)
18:41cilkay Aha, the second xmexample2 has the syntax
18:41cilkay It's just a Python list.
18:41cilkay actually, dictionary
18:42newsham looks like a list to me.
18:43MarkWilliamso| cilkay: the disks bit is a list. the colons are inside the strings and just get parsed in the tools.
18:44dhendrix Hello. I was wondering if anyone has an example filling in a multicall_entry_t and using do_multicall(). I can't seem to find any such examples by googling around...
18:45cilkay The colons make it look like a dictionary, though we don't have the {} denote a dict. I hadn't paid attention to the single quotes around the list elements.
18:46cdub dhendrix: did you grep for HYPERVISOR_multicall() and look at the various call sites?
18:46MarkWilliamso| dhendrix: I think thery're used in the block / net backend drivers
18:47dhendrix cdub: I've just looked for multicall(), I'll check for HYPERVISOR_multicall() as well.
18:47dhendrix MarkWilliamson: Thanks! I'll check those out and see if I can find some useful info from there.
18:48MarkWilliamso| dhendrix: OTOH, I may just be remembering batched MMU ops, which may work using a different mechanism (???)
18:49cdub MarkWilliamson: my grep says you're right
18:49MarkWilliamso| cdub: ah, good.
18:55--- ---> rusty [~rusty@ppp61-60.lns1.cbr1.internode.on.net] has joined #xen
18:57cdub rusty: morning ;-)
19:01rusty cdub: hi!
19:04cdub rusty: you going to montreal and ottawa?
19:10cilkay LVM backed vbd has booted!
19:10rusty cdub: Yeah, and I'll be in portland for Mon Tues too...
19:10cilkay I'm still getting the "respawning too fast message" and some module deps. iptables is also complaining.
19:11cilkay ... in domU
19:11cdub rusty: ah, fun ;-)
19:11cilkay So, Mandriva 2005 LE dom0, Mandriva 2005 LE domU with all but / as a LVM vbd
19:12MarkWilliamso| cilkay: It's working? Excellent!
19:12cdub rusty: we'll have to grab some cold ones
19:14rusty cdub: I'll need them after a week in Austin... it's always fun to catch up with the Oregon crowd
19:14cdub rusty: austin. oy! you're doing to full tour
19:15rusty Yeah, once you cross the pond it's hard to avoid "just one more" stop.
19:15cdub hehe
19:16cilkay Yes, it's working. Thank you very much to everyone who helped, and of course the folks who developed Xen in the first place.
19:19cilkay If I want to eliminate the errors on startup with missing modules, I guess I'll have to build a custom kernel.
19:23MarkWilliamso| cilkay: depends what modules they are
19:23MarkWilliamso| cilkay: if they're in the Xen distribution by default you can just install them into the guest's filesystem
19:23cilkay I copied the entire xenU modules directory to the guest
19:24cilkay ... after doing a depmod on it
19:24MarkWilliamso| ah ok, then you probably need to build some more :-)
19:24MarkWilliamso| There are certain startup scripts you'll want to disable in order to get rid of some errors.
19:27--- <<-- chrish01 [~chrish01@69.90.131.10] has quit (Quit: Leaving)
19:33--- <<-- aliguori [~anthony@cpe-70-112-81-91.austin.res.rr.com] has quit (Quit: Leaving)
19:40dhendrix rusty: David Hendricks from the speed team in Beaverton here. We don't mean to be a thorn in your side, but did you get our e-mail earlier cc'd to the vmtools-speedteam mailing list?
19:40rusty dhendrix: Just replied to it.
19:40dhendrix rusty: Ok, checking e-mail...
19:45MarkWilliamso| does anybody else here have trouble with domU reboots in the unstable tree?
19:45--- <<-- rusty [~rusty@ppp61-60.lns1.cbr1.internode.on.net] has quit (Quit: Client exiting)
19:45MarkWilliamso| Doesn't work for me but there's a reasonable probability of having broken it myself.
19:46dhendrix MarkWilliamson: Hmm, I did have a weird problem with sudden domU death syndrome a while ago, but that was after a bk pull so lord only knows what got broken there...
19:47MarkWilliamso| dhendrix: they're not dying unexpectedly, it's just that they're not coming back :-)
19:48MarkWilliamso| The code for restarting a domain still seems to be there but I guess it might be broken at the moment. I'm building a vanilla tree so we'll see...
19:58--- <<-- thelsdj [~thelsdj@24-117-42-27.cpe.cableone.net] has quit (Read error: Connection reset by peer)
20:03--- ---> thelsdj [~thelsdj@24-117-42-27.cpe.cableone.net] has joined #xen
20:10cilkay The "respawning too fast" messages fixed.
20:10--- <<-- tessier [~treed@222.253.78.207] has quit (Ping timeout: 480 seconds)
20:10cilkay Just comment out the VT2 through 6 in inittab.
20:11--- <<-- deac [~deac@xdsl-213-196-198-138.netcologne.de] has quit (Ping timeout: 480 seconds)
20:17--- <<-- malahal [~malahal@bi01p1.co.us.ibm.com] has quit (Quit: Leaving)
20:22MarkWilliamso| cilkay: nice one
20:22MarkWilliamso| good night all.
20:22--- <--- MarkWilliamson [~MarkW@maw48.kings.cam.ac.uk] has left #xen (Kopete 0.10 : http://kopete.kde.org)
20:22cilkay g'nite
20:22--- ---> deac [~deac@xdsl-213-196-202-104.netcologne.de] has joined #xen
20:23cilkay MarkWilliamson, what I'd really like to do is create a drak tool for Xen for Mandriva.
20:23cilkay That would be pretty sweet.
20:44--- ---> tessier [~treed@222.253.84.134] has joined #xen
21:52newsham # xm list
21:52newsham Name Id Mem(MB) CPU VCPU(s) State Time(s) Console
21:52newsham Domain-0 0 123 0 1 r---- 223.4 s:b731/p:33fc
21:52newsham whats the extra garbage at the end? (s:.. p:..)
21:53newsham (I get that after I restart xend)
22:01newsham (called ssidref in the code)
22:02--- <<-- drbyte [~cc2@c220-237-62-156.eburwd8.vic.optusnet.com.au] has quit (Quit: Leaving)
22:04newsham anyone alive who knows a little about libxc and xend? I need to debug something and have some questions
22:59--- <--- VS_ChanLog [~stats@ns.theshore.net] has left #xen (Rotating Logs)
22:59--- ---> VS_ChanLog [~stats@ns.theshore.net] has joined #xen
23:14--- ---> chrish01 [~chrish01@69.90.131.10] has joined #xen
23:27--- <<-- thelsdj [~thelsdj@24-117-42-27.cpe.cableone.net] has quit (Remote host closed the connection)
23:27--- ---> thelsdj [~thelsdj@24-117-42-27.cpe.cableone.net] has joined #xen
23:41--- ---> nivhome [~niv@c-67-171-167-143.hsd1.or.comcast.net] has joined #xen
23:55--- ---> rusty [~rusty@bh02i525f01.au.ibm.com] has joined #xen
---Logclosed Thu Jun 23 00:00:51 2005