| --- | Log | opened Thu Nov 08 00:00:23 2007 |
| 00:07 | |-| | setuid [~setuid@65.172.152.98] has joined #xen |
| 00:26 | |-| | hollisb [~hollisb@user-0vvdaf6.cable.mindspring.com] has quit [Quit: leaving] |
| 00:49 | <icarus901> | setuid, there? |
| 00:57 | <setuid> | I am |
| 00:57 | |-| | segher [~segher@82-217-247-28.cable.quicknet.nl] has quit [Ping timeout: 480 seconds] |
| 00:59 | <icarus901> | setuid, issues on freenode |
| 00:59 | <icarus901> | did you receive any messages from me |
| 01:00 | <setuid> | On Freenode? No, just the bit about the image itself |
| 01:00 | <icarus901> | hm |
| 01:00 | <setuid> | I replied with: <setuid> What I have, is a plain-vanilla AMD64/4600+ machine w/2gb RAM |
| 01:00 | <icarus901> | ah ok |
| 01:01 | <icarus901> | i said i can supply the image if you wish |
| 01:01 | <setuid> | I'm going to do a completely clean host reinstall, and do 3.0.3 with my notes and see if maybe something went wacky in the host |
| 01:01 | <setuid> | Or I can try 3.0.4 on the current install, and see if the problems just "Go Away" |
| 01:01 | <setuid> | How big is the image? |
| 01:02 | <icarus901> | well i used a sparse file |
| 01:02 | <icarus901> | so 24gb, but compressible to like |
| 01:02 | <icarus901> | 200mb |
| 01:02 | <setuid> | Hrm.. How did you create the sparse file? |
| 01:03 | <setuid> | I used dd and /dev/zero iirc. |
| 01:03 | <icarus901> | same |
| 01:03 | <setuid> | Ok |
| 01:03 | <icarus901> | just tell it to seek |
| 01:03 | <setuid> | What did you do differently to get it working? |
| 01:04 | <icarus901> | dd if=/dev/zero of=export.file bs=1024k seek=$((1024*NUM_GIGS)) count=1 |
| 01:04 | <icarus901> | nothing |
| 01:04 | <icarus901> | im going to investigate something though |
| 01:04 | <setuid> | Neat, never used that syntax before |
| 01:04 | <icarus901> | the $(()) is just bash math of course |
| 01:04 | <icarus901> | but for convenience |
| 01:04 | <setuid> | Of course |
| 01:05 | |-| | segher [~segher@82-217-247-28.cable.quicknet.nl] has joined #xen |
| 01:05 | <icarus901> | the bs should really be tweaked a bit or you may encounter some nasty performance issues with respect to cache alignment |
| 01:05 | <icarus901> | but 1024 is nice and convenient |
| 01:05 | <icarus901> | anyway |
| 01:06 | <icarus901> | im going to try something that might increase chances of success |
| 01:06 | <setuid> | Should I set up the host with lvm>? |
| 01:08 | <icarus901> | your choice |
| 01:08 | <icarus901> | i prefer it |
| 01:08 | <icarus901> | or just install normally |
| 01:08 | <icarus901> | and have an LVM disk or partition |
| 01:08 | <icarus901> | which you create after install |
| 01:15 | <setuid> | I'm reinstalling with a clean LVM host now on the metal |
| 01:15 | <icarus901> | ok |
| 01:27 | <setuid> | Clean lvm install on bare metal done, installing xen tools/kernels/etc. now |
| 01:31 | |-| | sputhenp [~sputhenp@202.80.58.210] has joined #xen |
| 01:37 | <icarus901> | setuid, assuming you encounter the problem again, kill the vm, then restart -- BUT attach strace -p THEPID -o somefile immediately |
| 01:38 | <icarus901> | also, if youre using sdl, use vnc just for now |
| 01:38 | <setuid> | I don't use either |
| 01:39 | <icarus901> | in what way are you accessing the framebuffer of the vm? |
| 01:43 | <setuid> | I just ssh to the host, then launch xm create from there, and it pops up a window on my local machine using X11 forwarding (I assume) |
| 01:44 | |-| | lymeca| [~lymeca@tor-irc.dnsbl.oftc.net] has quit [Ping timeout: 480 seconds] |
| 01:45 | |-| | lymeca| [~lymeca@tor-irc.dnsbl.oftc.net] has joined #xen |
| 01:47 | <icarus901> | setuid, that'd be sdl on the backend |
| 02:16 | <setuid> | ALmost there |
| 02:22 | <icarus901> | setuid, if/when the vm pegs the cpu, etc strace it for me for like 3 seconds and tell me what it says |
| 02:22 | <setuid> | Ok |
| 02:23 | <setuid> | Have to hack out the hard-coded ethernet device in the scripts, something weird in the qemu logs |
| 02:23 | <setuid> | /etc/xen/scripts/qemu-ifup: could not launch network script |
| 02:23 | <setuid> | Could not initialize device 'tap' |
| 02:26 | <icarus901> | setuid, do you have the tun/tap module installed |
| 02:26 | <setuid> | Will in a sec. |
| 02:28 | <setuid> | icarus901, Thought I had it... what provides tun/tap? |
| 02:30 | <icarus901> | setuid, vtun i believe |
| 02:30 | <setuid> | Trying that now |
| 02:31 | |-| | mastermind [~mastermin@mastermind.kaltenbrunner.cc] has quit [Remote host closed the connection] |
| 02:36 | <setuid> | HAd to install xorg |
| 02:44 | <setuid> | Somewhat closer... |
| 02:45 | <setuid> | icarus901, Is this what you meant? NUM_GIGS=22 dd if=/dev/zero of=disk.img bs=1024k seek=$((1024*NUM_GIGS)) count=22000 |
| 02:46 | <icarus901> | no to do sparse, you need to have count=1 |
| 02:46 | <icarus901> | but otherwise yes |
| 02:46 | <setuid> | When I do that, it never creates a disk image large enough |
| 02:46 | <setuid> | So when I boot the guest, it sees it as a 0.10G image |
| 02:47 | <icarus901> | this is the exact command i used dd if=/dev/zero of=export.file bs=1024k seek=$((23*1024)) count=1 |
| 02:48 | <setuid> | ah-HA! Bash isn't expanding it in my version |
| 02:48 | <icarus901> | could be |
| 02:49 | <setuid> | Testing again now |
| 02:49 | <setuid> | it books the guest, which is good... ultra-minimal install on the host |
| 02:49 | <setuid> | Trying to keep the host-level mods as minimal as possible |
| 02:49 | <icarus901> | id consider a full xorg install a little overkill haha |
| 02:50 | <icarus901> | oy the expansion wasnt happening thanks to NUM_GIGS not having a $ before it |
| 02:50 | <setuid> | I didn't install the full kit, but the last build, I got gdm and friends installed, so I could try to test this graphical installer "natively" (since it was cutting me off when doing it over the network) |
| 02:50 | <icarus901> | sorry |
| 02:51 | <icarus901> | ah i see |
| 02:51 | <icarus901> | by the way, you can make a second bridge attached to a dummy interface which goes nowhere |
| 02:51 | <icarus901> | that could make things easier |
| 02:51 | <setuid> | I should have caught that, I do a lot of bash... I'm just over-tired |
| 02:51 | <icarus901> | likewise |
| 02:52 | <setuid> | I want to get this working on my end, before I crash for the night, and have to bring the little one to daycare in 5 hours |
| 03:00 | <setuid> | debuntangle:/var/xen/domains/vm01.debuntangle.gnu-designs.com# Timeout, server not responding. |
| 03:01 | <setuid> | dead |
| 03:01 | <setuid> | So its reproducable |
| 03:01 | <setuid> | I'll give 3.1.0 a try |
| 03:02 | <icarus901> | setuid, wait a moment |
| 03:02 | <icarus901> | tell me what strace says |
| 03:04 | <setuid> | Capturing strace output now |
| 03:06 | <icarus901> | a few seconds should be sufficient |
| 03:10 | <setuid> | Ok, after about 300 network bus resets, it finally disconnected for sure |
| 03:10 | <setuid> | running strace on the pid got it to stay up longer, but it eventually died, whacking out all of my open terminals in the process |
| 03:10 | <setuid> | Now I can't ssh to the host at all |
| 03:10 | <icarus901> | can you pastebin the strace |
| 03:13 | <setuid> | Yep, dom0's network was taken out... and isn't coming back |
| 03:13 | |-| | setuid [~setuid@65.172.152.98] has quit [Remote host closed the connection] |
| 03:25 | |-| | nantes_geek [~ygcheny@sma-themericourt.rain.fr] has joined #xen |
| 03:27 | |-| | nantes_geek [~ygcheny@sma-themericourt.rain.fr] has quit [] |
| 03:29 | |-| | nantes_geek [~ygcheny@sma-themericourt.rain.fr] has joined #xen |
| 04:05 | |-| | nantes_geek [~ygcheny@sma-themericourt.rain.fr] has quit [Ping timeout: 480 seconds] |
| 04:06 | |-| | nantes_geek [~ygcheny@sma-themericourt.rain.fr] has joined #xen |
| 04:24 | |-| | boto [~ehabkost@189.32.40.212] has joined #xen |
| 04:34 | |-| | am88b [siim@tux.linux.ee] has joined #xen |
| 04:52 | <nantes_geek> | hi ! |
| 04:52 | <nantes_geek> | i have got a problem to boot an hvm with pxe |
| 04:53 | <nantes_geek> | i have : Nov 8 10:47:09 xen0 kernel: device vif43.0 left promiscuous mode |
| 04:55 | <nantes_geek> | qemu: invalid boot device in 'n' |
| 05:25 | |-| | sputhenp [~sputhenp@202.80.58.210] has quit [Quit: Leaving] |
| 05:26 | |-| | bernarde [~bernarde@143.106.167.152] has joined #xen |
| 05:45 | |-| | sputhenp [~sputhenp@202.80.58.210] has joined #xen |
| 05:47 | |-| | nantes_geek [~ygcheny@sma-themericourt.rain.fr] has quit [Ping timeout: 480 seconds] |
| 05:56 | |-| | sputhenp [~sputhenp@202.80.58.210] has quit [Quit: Leaving] |
| 05:56 | |-| | sputhenp [~sputhenp@202.80.58.210] has joined #xen |
| 05:57 | |-| | quintela [~quintela@trasno.trasno.org] has joined #xen |
| 07:04 | |-| | jwb [~jwboyer@68-115-98-121.dhcp.roch.mn.charter.com] has joined #xen |
| 07:05 | |-| | jwb [~jwboyer@68-115-98-121.dhcp.roch.mn.charter.com] has left #xen [] |
| 07:05 | |-| | beth_ changed nick to beth |
| 07:12 | |-| | sputhenp [~sputhenp@202.80.58.210] has quit [Quit: Leaving] |
| 07:18 | |-| | mrj [~mrj@sca-ea-fw-1.Sun.COM] has joined #xen |
| 07:27 | |-| | beth [~beth@ool-45777805.dyn.optonline.net] has quit [Remote host closed the connection] |
| 07:34 | |-| | beth [~beth@ool-45777805.dyn.optonline.net] has joined #xen |
| 07:47 | |-| | infernix [nix@spirit.infernix.net] has quit [Quit: Leaving] |
| 08:17 | |-| | nantes_geek [~ygcheny@sma-themericourt.rain.fr] has joined #xen |
| 08:18 | |-| | mrj [~mrj@sca-ea-fw-1.Sun.COM] has quit [Ping timeout: 480 seconds] |
| 08:18 | |-| | sputhenp [~sputhenp@202.80.58.210] has joined #xen |
| 08:27 | |-| | agriffis [~agriffis@ata01cs602.americas.hp.net] has quit [Quit: leaving] |
| 08:29 | |-| | agriffis [~agriffis@cor000cs01.americas.hp.net] has joined #xen |
| 08:33 | <spuk-> | is ti possible to make the console larger? right now I get a 80x25 console inside a 127x42 xterm, looks awkward.. |
| 08:35 | <spuk-> | oh, just exporting COLUMNS and LINES manually and re-executing bash works.. :-p |
| 08:36 | |-| | boto [~ehabkost@189.32.40.212] has quit [Ping timeout: 480 seconds] |
| 08:38 | |-| | mrj [~mrj@c-75-69-174-59.hsd1.nh.comcast.net] has joined #xen |
| 08:39 | <spuk-> | i guess the thing would be to make the linux console code know the outer terminal lines/columns ? |
| 08:40 | |-| | infernix [nix@spirit.infernix.net] has joined #xen |
| 08:42 | |-| | mrj_ [~mrj@sca-ea-fw-1.Sun.COM] has joined #xen |
| 08:48 | |-| | mrj [~mrj@c-75-69-174-59.hsd1.nh.comcast.net] has quit [Ping timeout: 480 seconds] |
| 09:14 | |-| | clalance [~clalance@nat-pool-bos.redhat.com] has joined #xen |
| 09:40 | |-| | sputhenp [~sputhenp@202.80.58.210] has quit [Quit: Leaving] |
| 09:44 | |-| | sputhenp [~sputhenp@202.80.58.210] has joined #xen |
| 10:02 | |-| | _gh_ [~gerrit@c-67-169-199-103.hsd1.or.comcast.net] has joined #xen |
| 10:21 | |-| | boto [~ehabkost@189.32.40.212] has joined #xen |
| 10:39 | |-| | hbaum [~hbaum@bi01p1.co.us.ibm.com] has joined #xen |
| 10:50 | |-| | rharper [~rharper@bi01p1.co.us.ibm.com] has joined #xen |
| 11:00 | |-| | setuid [~setuid@65.172.152.98] has joined #xen |
| 11:08 | |-| | sputhenp [~sputhenp@202.80.58.210] has quit [Quit: Leaving] |
| 11:15 | |-| | ahs3 [~ahs3@66.37.132.58] has joined #xen |
| 11:33 | |-| | danp1 [~berrange@nat-pool-bos.redhat.com] has joined #xen |
| 11:36 | |-| | danpb [~berrange@nat-pool-bos.redhat.com] has quit [Quit: Leaving.] |
| 11:38 | |-| | mdday [~mdday@cpe-075-189-153-207.nc.res.rr.com] has joined #xen |
| 11:39 | |-| | hollisb [~hollisb@bi01p1.co.us.ibm.com] has joined #xen |
| 11:40 | |-| | KriP [kp@ip2-53.vejen-net.dk] has joined #xen |
| 12:03 | |-| | mastermind [~mastermin@mastermind.kaltenbrunner.cc] has joined #xen |
| 12:42 | |-| | boto_ [~ehabkost@189.32.40.212] has joined #xen |
| 12:42 | |-| | mastermind [~mastermin@mastermind.kaltenbrunner.cc] has quit [Ping timeout: 480 seconds] |
| 12:47 | |-| | boto [~ehabkost@189.32.40.212] has quit [Ping timeout: 480 seconds] |
| 12:50 | |-| | mastermind [~mastermin@mastermind.kaltenbrunner.cc] has joined #xen |
| 13:24 | <icarus901> | icblenke, geoff_k_ why on earth are things so unstable lately |
| 13:27 | <icblenke> | 1. Acquisition of XenSource by Citrix, 2. Various vendor specific forks, 3. Merging into the mainline kernel, 4. lguest/virtops/etc |
| 13:28 | <icblenke> | that's what I'm seeing from the outside looking in, anyway. |
| 13:29 | <icarus901> | icblenke, i dont disagree there |
| 13:29 | <icarus901> | however i was speaking with respect to irc |
| 13:29 | <icarus901> | netsplits |
| 13:29 | |-| | arun [~arun@yktgi01e0-s5.watson.ibm.com] has joined #xen |
| 13:29 | <icblenke> | ah, heh. I have /ignore * crap, so I don't see them. |
| 13:29 | <icarus901> | ah |
| 13:30 | <icarus901> | icblenke, i honestly dont know if youve received any messages from me lately as a result |
| 13:30 | <icarus901> | have you, by any chance? |
| 13:30 | <icblenke> | not that I'm aware of. |
| 13:32 | [~] | setuid decides to stay here, to maintain continuity |
| 13:32 | <icarus901> | setuid, please do |
| 13:32 | <setuid> | icarus901, I'm trying a 3.1.0 tree now, trying to pick my way through 'make dist' to make sure it doesn't go breaking things |
| 13:32 | <icarus901> | icblenke, primarily ive asked about bridging and flipping interfaces |
| 13:33 | <icarus901> | you mentioned them a long while ago, and ive been asked several times about such problems |
| 13:36 | <icblenke> | we still see it. |
| 13:36 | <icblenke> | but I've modified my vif-bridge script to correct the flipping when it happens. |
| 13:36 | <icarus901> | icblenke, what is the root cause |
| 13:37 | <icarus901> | (i either pass a pci dev through to a domU or use a single interface, so i have yet to see this personally) |
| 13:38 | [~] | murb is very happy that the dom0 loopback net devices will be dying in 3.2 |
| 13:38 | <murb> | s/dying/not required/ |
| 13:40 | <setuid> | flipping? |
| 13:46 | <icarus901> | setuid, occasionally, a domU with two interfaces bound to two bridges will swap |
| 13:46 | <icarus901> | which as you might imagine is rather annoying |
| 13:46 | <icarus901> | icblenke, willing to pastebin your vif-bridge? |
| 13:47 | |-| | riel [U2FsdGVkX1@74.92.59.66] has left #xen [Client exiting] |
| 13:47 | <icblenke> | I don't read the xen config file to deduce what has flipped, I generate the xen domu config file from my own configuration tree (filesystem based) which is also what my modified vif-bridge checks. |
| 13:47 | <setuid> | icarus901, Could that be something related to what I'm still fighting over here? |
| 13:48 | <icarus901> | setuid, unlikely |
| 13:48 | <icblenke> | I don't think your problem is related to this, no. |
| 13:48 | <setuid> | Ok |
| 13:48 | <icarus901> | icblenke, that's certainly understandable |
| 13:49 | <icarus901> | i'm going to reproduce and poke at it a bit |
| 13:49 | <murb> | icarus901: i've never seen that.. |
| 13:50 | <icarus901> | murb, neither have i. but apparently its common enough to be a question often asked |
| 13:50 | <setuid> | icarus901, Should I just do a 'make dist' and install? |
| 13:50 | <setuid> | Or should I try to convert that into a .deb? |
| 13:55 | <geoff_k_> | problems again i see jsut got back form a quick shopping trip |
| 13:56 | <geoff_k_> | seems to be that simmons host all the time |
| 14:14 | |-| | bernarde [~bernarde@143.106.167.152] has quit [Remote host closed the connection] |
| 14:20 | <icarus901> | setuid, your choice |
| 14:21 | <icarus901> | setuid, just take care to be sure that the older xen installation does not interfere |
| 14:22 | <setuid> | I guess I don't know how I'd do that... I'll have to rewrite the init script, the rc scripts and symlinks, rebuild the kernel against that new xen version, overwrite the binaries, etc. |
| 14:22 | <setuid> | it'll take a few days to get that all figured out |
| 14:23 | <murb> | setuid: you will have to do nearly nothing as it has already been done. |
| 14:23 | <setuid> | oh? |
| 14:24 | <icarus901> | setuid, you could just remove 3.0.3 via apt.. |
| 14:28 | <setuid> | Ripping it all out now |
| 14:29 | |-| | mdday_ [~mdday@cpe-075-189-153-207.nc.res.rr.com] has joined #xen |
| 14:29 | <setuid> | It tangles itself so tightly to the system, I had no idea it was such a vertical climb to get Linux running in Xen |
| 14:29 | <icarus901> | it isnt |
| 14:29 | <icarus901> | its the rather too-old debian packages |
| 14:29 | <danp1> | icarus901: it is |
| 14:30 | <icarus901> | which danp1 |
| 14:30 | <icarus901> | debian or the vertical climb |
| 14:30 | <danp1> | icarus901: the amount of infrastructure you need to install to get Xen running is insane |
| 14:30 | <icarus901> | ive had astounding luck then |
| 14:30 | <danp1> | icarus901: i've got prototype code which makes Xen as simple as KVM - eg merely spawn a qemu binary with neccessary args |
| 14:31 | <icarus901> | ive had literally one problem since 3.0.2 |
| 14:31 | <icarus901> | none in the 2.x.x line |
| 14:31 | <setuid> | I'm into my 3rd day burned on this |
| 14:31 | <icarus901> | either ive been profoundly lucky or my patience is unusual danp1 ? |
| 14:32 | <danp1> | icarus901: well you've been using it long enough to understand it i guess |
| 14:32 | <icarus901> | setuid, yes your issue is rather damning |
| 14:32 | <icarus901> | i suppose danp1 |
| 14:32 | <icarus901> | i do not by any means disagree that the shift in thinking is substantia |
| 14:32 | <icarus901> | for younger sysadmins, its a trial by fire |
| 14:32 | |-| | mdday [~mdday@cpe-075-189-153-207.nc.res.rr.com] has quit [Ping timeout: 480 seconds] |
| 14:34 | <icarus901> | setuid, you've clearly been deep in this stuff for a long time so i'm not putting you into that group |
| 14:34 | |-| | tof [~tof@195.95.224.45] has joined #xen |
| 14:34 | <icarus901> | you're experiencing a legitimate bug |
| 14:35 | <setuid> | Well, I'm green with regard to xen, just need to translate the nuances in xen with what I have implemented using other virtualization solutions |
| 14:36 | |-| | _mcp [~hightower@wolk-project.de] has joined #xen |
| 14:36 | <icarus901> | setuid, i was asking about the strace output because occasionally a severed connection to the qemu-backed HVM framebuffer causes the qemu-dm process to go batshit insane checking time |
| 14:36 | |-| | SR|Ultra_ [~lec9@falcon.le.ac.uk] has joined #xen |
| 14:36 | <icarus901> | danp1, glad to see fc8 will have virsh via ssl to other nodes |
| 14:36 | |-| | Netsplit synthon.oftc.net <-> solenoid.oftc.net quits: spuk-, tof_, ag-, C14r, SR|Ultra, mcp |
| 14:36 | |-| | Netsplit over, joins: spuk- |
| 14:36 | |-| | _mcp changed nick to mcp |
| 14:36 | <setuid> | I did run a strace last night, but it still died out... and I had to power the box off to work on another project (not enough hardware to go around), so I lost that strace output. If 3.1.0 fails, I'll try again and strace it. |
| 14:37 | |-| | Netsplit over, joins: ag- |
| 14:37 | <danp1> | icarus901: its already available as an update to f7 |
| 14:37 | <icarus901> | ah, very good, i was unaware |
| 14:38 | <danp1> | we'll be adding kerberos too soon |
| 14:38 | <danp1> | and digest-md5 for plain username/passwd hashes |
| 14:38 | <icarus901> | ldap? |
| 14:38 | <danp1> | eventually |
| 14:38 | <danp1> | well for auth, basically anything SASL supports (which includes ldap) |
| 14:39 | <setuid> | ldap auth into the xen guests? I plan on doing that as my first step |
| 14:39 | <danp1> | but for ACLs we need to implement ldap directly |
| 14:39 | <setuid> | I already have that modality working in vmware |
| 14:39 | <icarus901> | setuid, for guest management |
| 14:39 | <icarus901> | external to the guests |
| 14:39 | <setuid> | host has the ldap server, all guests use ldap to auth against the host, and guest home directories are nfs-mounted to the host's $HOME |
| 14:39 | |-| | C14r [~C14r@h58173.serverkompetenz.net] has joined #xen |
| 14:40 | <icarus901> | setuid, you could frankly convert your vmware guests to xen |
| 14:41 | <icarus901> | without having to set everything up from scratch |
| 14:41 | <setuid> | Except brctl |
| 14:41 | <icarus901> | i'm hoping setuid that the issue will be resolved for you in 3.1 |
| 14:41 | <setuid> | Yep, its building now... takes a long time, even on this fast hardware |
| 14:42 | <setuid> | I should have done a 'make -j12 dist' ;) |
| 14:42 | <icarus901> | oh lord |
| 14:42 | <setuid> | distclean removed the kernel tarball too, probably not a good idea |
| 14:42 | <setuid> | Since rebuilding required a re-download of a 49M tarball ;) |
| 14:42 | <icarus901> | setuid, pastebin a verbose lspci and cpuinfo, etc |
| 14:42 | <icarus901> | i want to see if its possible your system has something funky |
| 14:45 | <setuid> | http://rafb.net/p/b30o8469.html |
| 14:48 | |-| | fuzzy [~fuzzy@64.62.120.120] has joined #xen |
| 14:57 | <brendan> | does anyone know much about tapdisk hacking? I have a driver I'm working on to buffer writes in memory until an explicit flush, but I'm not sure how to handle shutdown. the close method can't return busy, so it should flush before it returns. But my buffer sits on top of a base tapdisk, which just sets up async IO and then fires a tapdisk callback when it's done. I'm not sure what is the best way to intercept this message without returning from the close method |
| 14:58 | <setuid> | hrm, latest 3.1.0 doesn't seem to have any of the tools in xen-utils |
| 14:59 | <icarus901> | setuid, which tools |
| 15:00 | <setuid> | xen-create-image is one so far... but let me see if this thing works with my vm |
| 15:01 | <icarus901> | setuid, those are not xen utilities, those are provided by xen-tools which is a package developed by steve kemp, an individual with (as far as i know) no affiliation with citrix/xensource |
| 15:01 | <setuid> | ahhh |
| 15:01 | <setuid> | I thought it was one of those packages Debian splits out from the core upstream package |
| 15:02 | <icarus901> | while that is their tendency, nope |
| 15:31 | <icarus901> | setuid, status? |
| 15:32 | <setuid> | Just finished downloading the kernel |
| 15:32 | <setuid> | 15:23:29 (71.42 KB/s) - `linux-2.6.18.tar.bz2' saved [41863580/41863580] |
| 15:32 | <setuid> | _verrrrry_ slow to kernel.org these days |
| 15:36 | |-| | danpb [~berrange@nat-pool-bos.redhat.com] has joined #xen |
| 15:43 | |-| | danp1 [~berrange@nat-pool-bos.redhat.com] has left #xen [] |
| 15:44 | <murb> | setuid: use a mirror. |
| 15:44 | <setuid> | I tried... its ok, its building dist again now |
| 15:54 | |-| | danpb [~berrange@nat-pool-bos.redhat.com] has quit [Quit: Leaving.] |
| 15:58 | |-| | danpb [~berrange@nat-pool-bos.redhat.com] has joined #xen |
| 16:17 | <setuid> | heh, found a typo in the README for 3.1.0 |
| 16:17 | <setuid> | module /boot/vmlinuz-2.6-xen root=<root-dev> ro console=tty0 |
| 16:17 | <setuid> | module /boot/initrd-2.6-xen.img |
| 16:17 | <setuid> | Second line should be: |
| 16:18 | <setuid> | initrd /boot/initrd-2.6-xen.img |
| 16:18 | <murb> | no it shouldn't. |
| 16:19 | <murb> | the above it so correct as far as it goes. |
| 16:19 | <setuid> | Two module lines? |
| 16:19 | <setuid> | I've never seen an initrd work as a module before |
| 16:19 | <hensema> | that's the way our servers are setup too |
| 16:28 | <murb> | setuid: have you ever used grub with a Multiboot-compliant kernel? |
| 16:35 | |-| | boto_ [~ehabkost@189.32.40.212] has quit [Remote host closed the connection] |
| 17:04 | |-| | clalance [~clalance@nat-pool-bos.redhat.com] has quit [Quit: Leaving] |
| 17:05 | <setuid> | murb, No |
| 17:06 | <setuid> | All of the virtualization I've ever done was independent of the host's boot process and kernel |
| 17:08 | <icarus901> | setuid, technically, the host is a virtual machine too |
| 17:08 | <setuid> | Looks like the kernel that 3.1.0 builds is a no-op |
| 17:08 | <setuid> | So I'm screwed. |
| 17:09 | <setuid> | Can't get my vm working with 3.0.3, no 3.0.4 deb for this version of Debian, and 3.1.0 isn't booting... just keeps rebooting the box when it unpacks the kernel |
| 17:10 | <setuid> | I don't want to tell the client to create their guests in vmware and convert them to xen vms |
| 17:11 | |-| | mrj_ [~mrj@sca-ea-fw-1.Sun.COM] has quit [Quit: mrj_] |
| 17:11 | [~] | setuid grumbles... I have a client who owes me $3k for some LDAP work, and they haven't paid up yet. |
| 17:13 | <murb> | setuid: :( yeah you should have invoiced in euro. |
| 17:14 | <murb> | setuid: so the stock debs don't work at all? |
| 17:16 | <setuid> | murb, I haven't identified the source of the problem yet... trying to see if the vm configuration I have works with 3.0.4, because it doesn't work on 3.0.3... base vm, booted an iso, installed the software, rebooted the vm, it comes back up and does the post-install setup, then about 90% of the way through, the _guest_, knocks out all _host_ networking. |
| 17:17 | <setuid> | So all shells and apps running on the box are cut, and then the networking comes back up, but the vm is dead... and since xconsole doesn't support hvm, I can't resume it. |
| 17:17 | <setuid> | So I have to destroy and start over. |
| 17:17 | <setuid> | Lather. Rinse. Repeat. |
| 17:17 | <setuid> | Happens every single time. |
| 17:18 | |-| | KriP [kp@ip2-53.vejen-net.dk] has quit [Quit: Leaving] |
| 17:19 | <setuid> | btw, the two module lines are invalid, Debian removes the second one |
| 17:19 | <danpb> | setuid: the 2 modules lines works just fine & as expected |
| 17:20 | <danpb> | setuid: i'm booted up in such a config right now |
| 17:20 | <setuid> | danpb, Except on Debian, because their grub strips the second one out. |
| 17:20 | <danpb> | well they've got a broken grub then |
| 17:20 | <setuid> | I can add the line, run update-grub, and look at menu.lst, and its gone |
| 17:20 | <setuid> | I can add the line, reboot, and when it comes back up, its gone |
| 17:21 | <murb> | setuid: how does it remove it? |
| 17:22 | <murb> | setuid: are you using update-grub? |
| 17:24 | <setuid> | Looks like 3.1.0 doesn't use initrd |
| 17:24 | <murb> | setuid: I have here: |
| 17:24 | <murb> | kernel /boot/xen-3.1.2-rc2.gz dom0_mem=524288 |
| 17:24 | <murb> | module /boot/xen0-linux-2.6.18xen0 root=/dev/whatever |
| 17:24 | <danpb> | setuid: sure it does |
| 17:24 | <murb> | module /boot/initrd.img-2.6.18xen0 |
| 17:24 | <murb> | and it works fine |
| 17:24 | <setuid> | danpb, Ok, 'make dist' does not create one |
| 17:25 | <danpb> | initrd's are specific to your particular install |
| 17:25 | <setuid> | A lot of this seems so... "manual", but then again, that's all we had 10+ years ago, so I don't mind digging in and doing it. |
| 17:25 | <danpb> | its not appropriate for make dist to make one up |
| 17:26 | <danpb> | and every distro has a different way initrd's are put together so Xen can't satisfy all these approaches |
| 17:26 | <setuid> | Is it normal to take > 3 days to get xen configured and a single guest installed and working? |
| 17:26 | <murb> | setuid: try /usr/sbin/update-initramfs -k kernelversion -c |
| 17:26 | <murb> | setuid: no |
| 17:26 | |-| | hollisb [~hollisb@bi01p1.co.us.ibm.com] has quit [Quit: leaving] |
| 17:26 | <setuid> | I'm on Day #3, and my guest still doesn't complete its install without wiping out the host's network subsystem |
| 17:27 | <murb> | setuid: maybe you should hire someone to do it for you.. |
| 17:27 | <setuid> | And with no way to resume after the guest knocks out networking to the box, its a dead-end. |
| 17:27 | <setuid> | murb, I'm confident enough to be able to do it, but it appears that the errors lie either with the distro's udev/hotplug scripts, or within xen itself. |
| 17:28 | <setuid> | I've successfully reproduced the failure on 3 separate physical machines, all built clean from the ground up with Debian's netinst bare install + xen 3.0.3 |
| 17:29 | <murb> | setuid: well i installed xen 3.1.2-rc2 the other day from mercurial on a remote xen etch box without much trouble. |
| 17:29 | <murb> | ok it took me longer because the ethernet card in dom0 wasn't supported by the stock 2.6.18 kernel. |
| 17:29 | <setuid> | icarus901 suggested I try 3.0.4, but the download from the main site appears to be 3.1.0, so I tried building that up, but the kernel it builds does not boot, because it claims I have an invalid root= line (vfs panic). The root= line is the same as the other functioning kernels. |
| 17:30 | <murb> | setuid: did you build an initrd? |
| 17:30 | <murb> | setuid: do you normally include the support for your filesystem and disks in your kernel? |
| 17:30 | <murb> | the xen .config probably builds a very modular kernel. |
| 17:30 | <setuid> | Yes, but I didn't choose anything for this kernel, it was all bootstrapped for me |
| 17:31 | [~] | setuid wrote the kernel HOWTO, so that's something I'm very familiar with |
| 17:31 | <setuid> | Well, the second kernel howto anyway |
| 17:31 | <murb> | setuid: I really think you should start off with the stock debian xen and kernel packages. |
| 17:31 | <murb> | and try and get them to work. |
| 17:31 | <setuid> | Those work, except with my target guest's vm |
| 17:31 | <murb> | because now you are failing at a much ealier stage, this is not progress. |
| 17:32 | <murb> | what is the guest vm? |
| 17:32 | <setuid> | And I'm unclear whether the problem is 3.0.3, Debian's udev/hotplug scripts, or something inside the guest vm itself, that wipes out host networking |
| 17:32 | <setuid> | murb, Untangle(.com), it's a monitoring appliance. |
| 17:32 | <setuid> | It definitely does not work under 3.0.3, that much I've verified. |
| 17:32 | <murb> | setuid: sounds like the bridging |
| 17:32 | <setuid> | I'd like to use process of elimination to determine if it fails in 3.0.4 or 3.1.0 |
| 17:32 | <murb> | if you have spanning tree turned on the networking could go away for 30 seconds. |
| 17:33 | <setuid> | If it does, then the problem isn't xen, but udev/hotplug or the guest itself |
| 17:33 | <murb> | setuid: is the machine remote? |
| 17:33 | <setuid> | But a guest should NEVER wipe out networking in the host |
| 17:33 | <setuid> | The machine is downstairs in the lab, 1 floor below me |
| 17:33 | <murb> | the scripts have to do things to bridges etc that are on the host. |
| 17:35 | <setuid> | Is there some way to tell it NOT to do a bus reset during guest install? |
| 17:35 | <setuid> | It wipes out everything on the host, apache, mysql, transfers, everything. |
| 17:37 | <setuid> | I'm going to try this: http://unicorn.redone.hu/blog/?p=15 |
| 17:38 | <murb> | bsu reset? |
| 17:38 | <setuid> | yes |
| 17:38 | <setuid> | Here's what I am currently doing: |
| 17:38 | <setuid> | 1.) ssh to the host |
| 17:38 | <setuid> | 2.) Launch xm create /etc/xen/vm01-bleh.cfg |
| 17:39 | <murb> | are you trying to install an hvm guest? |
| 17:39 | <setuid> | 3.) Go through the graphical setup, reboot, go through the post-install setup |
| 17:39 | <murb> | ok what is in vm01-belh.cfg |
| 17:39 | <setuid> | 4.) While the setup is going, I open several other ssh shells to the host |
| 17:40 | <setuid> | 5.) When the guest post-install setup reaches about 90%, _all_ network traffic to the server is severed, guest vm window vanishes, all open ssh shells report a timeout, |
| 17:40 | <setuid> | I can immediately ssh back in, but now my guest vm is in an unretrievable state |
| 17:40 | <setuid> | Let me pastebot my .cfg |
| 17:41 | <murb> | during the installation is there lots of network traffic to the guest you are installing? |
| 17:42 | |-| | mdday_ changed nick to mday_away |
| 17:44 | <setuid> | Nope |
| 17:45 | <setuid> | Its relatively low-traffic, I'm just setting it up as a new server |
| 17:46 | <setuid> | Xen is most-definitely not a drop-in... have to go fetching third-party packages now to build hvmloader, etc. |
| 17:46 | <setuid> | I don't mean to rant, but man this is cryptic and complicated to get up and running |
| 17:49 | <setuid> | eek! The 'uninstall' target for xen source is *WAY* too broad |
| 17:49 | <setuid> | rm -rf //boot/*xen* |
| 17:49 | <setuid> | rm -rf //lib/modules/*xen* |
| 17:49 | <setuid> | And so on |
| 17:50 | <setuid> | That just wiped out some of the other non-xen(from source) things I had installed |
| 17:50 | <murb> | setuid: it does warn you! |
| 17:50 | <setuid> | What does? |
| 17:51 | <setuid> | make uninstall does no such thing |
| 17:51 | <murb> | @echo ' uninstall - attempt to remove installed Xen tools' |
| 17:51 | <murb> | @echo ' (use with extreme care!)' |
| 17:51 | <setuid> | http://rafb.net/p/DzOuld73.html |
| 17:51 | <setuid> | That must be in your devel version, it doesn't exist in the release |
| 17:52 | <setuid> | The problem is that it removes things that it didn't install, and that's *BAD* |
| 17:52 | <danpb> | make uninstall rules are always dangerous ....this is what package management is for |
| 17:52 | <murb> | hg tip # says Xen 3.1.0 final. |
| 17:52 | <setuid> | Of course |
| 17:53 | <setuid> | Can I use the hvmloader from 3.0.3 with 3.1.0? |
| 17:53 | <murb> | setuid: also don't run make install, make dist etc without atleast having a look inside the markfile. |
| 17:53 | <setuid> | Because I can't seem to find the right tools to build the firmware mated for 3.1.0 |
| 17:53 | <murb> | setuid: doubtful. |
| 17:53 | [~] | setuid bangs head on the desk |
| 17:54 | <murb> | setuid: grab the packages from testing. |
| 17:54 | <murb> | http://packages.debian.org/testing/xen-hypervisor-3.1-1-amd64 etc |
| 17:54 | <murb> | etc |
| 17:55 | <murb> | and if ness, rebuild them. |
| 17:55 | <setuid> | I'm going to upgrade the whole host to unstable, and try that |
| 17:55 | <murb> | no you are not. |
| 17:55 | <murb> | add in a deb-src line in your apt-sources |
| 17:56 | <murb> | http://packages.debian.org/source/lenny/xen-3 |
| 17:56 | <setuid> | I found somthnig: |
| 17:56 | <setuid> | deb-src http://packages.debianbase.de/sources/xen3.1 ./ |
| 17:56 | <murb> | the buliding depdancies don't look too bad. |
| 17:56 | |-| | arun [~arun@yktgi01e0-s5.watson.ibm.com] has quit [Ping timeout: 480 seconds] |
| 17:57 | <setuid> | Building software I can do, I do that in my sleep... but using w0nky third-party asm compilers to build and bootstrap firmware, is not something I'm used to doing |
| 17:57 | <murb> | qemu isn't *that* bad! |
| 17:59 | [~] | setuid tries to figure out a working version of deb-src line to go with that xen-3 repo |
| 18:00 | <murb> | deb-src ftp://ftp.de.debian.org/debian/ etch lenny # WFM |
| 18:00 | <setuid> | Not that one |
| 18:01 | |-| | mrj [~mrj@sca-ea-fw-1.Sun.COM] has joined #xen |
| 18:16 | |-| | aw [~awilliam@c-24-9-84-32.hsd1.co.comcast.net] has quit [Quit: Leaving] |
| 18:16 | |-| | aw [~awilliam@c-24-9-84-32.hsd1.co.comcast.net] has joined #xen |
| 18:16 | |-| | aw [~awilliam@c-24-9-84-32.hsd1.co.comcast.net] has quit [] |
| 18:18 | |-| | aw [~awilliam@c-24-9-84-32.hsd1.co.comcast.net] has joined #xen |
| 18:18 | <setuid> | How can I stop the guest from HUP'ing the host's network? |
| 18:18 | |-| | mrj [~mrj@sca-ea-fw-1.Sun.COM] has left #xen [] |
| 18:19 | <setuid> | There's a forum post on the untangle forums, indicating that someone has this working "under Xen 3.0.3 on Debian Etch (using the default Debian Xen packages)." |
| 18:32 | <setuid> | murb, You asked for my .cfg, here it is: http://rafb.net/p/jZtVry69.html |
| 18:38 | <murb> | why do you have two interfaces on the same bridge? |
| 18:38 | <setuid> | Because the guest needs to virtual NICs |
| 18:39 | <setuid> | And rkeene in #xen on Freenode suggested I use that syntax |
| 18:39 | <setuid> | The initial guest setup fails early on when it checks for two NICs |
| 18:39 | <murb> | if it is just a custom linux couldn't you run it as a PV guest? |
| 18:39 | <setuid> | Should I be using another approach? |
| 18:40 | <murb> | setuid: i'd use a dead simple configuration for the inital installation. |
| 18:42 | <setuid> | murb, that's as simple as I can get it |
| 18:44 | <murb> | does it refused to install with just one nic? |
| 18:44 | <murb> | and is your hardware known good? |
| 18:49 | <setuid> | Yes and yes |
| 18:49 | <setuid> | The servers are maybe 2-3 months old, and they ran 24x7 for web/mail/cvs/svn infrastructure until we migrated those services upstream |
| 18:51 | <setuid> | Is there some other way to represent two virtual NICs to the guest? |
| 18:52 | |-| | arun [~arun@ool-44c6b6b1.dyn.optonline.net] has joined #xen |
| 18:52 | <murb> | the vif = [] line should do the trick |
| 18:53 | <murb> | setuid: if the intention is that this appliance will be a firewall/router i stronly advise against doing it in an HVM domain wtihout pv driver.s |
| 18:54 | <setuid> | Starting UVM: Removing Stale PIDfile (file:2337 pid:2337) [....................................... |
| 18:54 | <setuid> | Seems to be stuck there in the guest, dots ticking away... and now I bet it'll whack the host's networking |
| 18:55 | <setuid> | murb, I thought running it in the hvm _was_ pv |
| 18:55 | <murb> | setuid: well if you have a loop in your topology some how. |
| 18:55 | <murb> | setuid: exactly not. |
| 18:55 | <murb> | hvm is full virtualsation that is why you have that strange firmware crap etc. |
| 18:55 | <setuid> | Is that necessary? |
| 18:55 | <setuid> | I mean, why do I need hvm, if I can do pv? |
| 18:55 | <murb> | setuid: if this appliance is based on linux you should be able to build a xenised kernel for it. |
| 18:56 | <murb> | setuid: if i was you i'd get it up and running say in vmware if you have that to hand or qemu, then migrate the filesystem to a raw image / lvm lv and install a xenised kernel. |
| 18:56 | <murb> | then you can forget all this hvm stuff. |
| 18:56 | <setuid> | Except that at some point, someone needs to do exactly what I'm doing, at client sites |
| 18:57 | <setuid> | i.e. install Debian, install Xen, install Untangle in Xen. |
| 18:57 | <murb> | setuid: maybe you can modify the untangle installer? |
| 18:57 | <murb> | so it uses a Xenised kernel to begin with? |
| 18:57 | <setuid> | I don't want to tell the clients that in order to get Xen guest vms working, they need to create it in vmware and convert it to a xen vm |
| 18:57 | <setuid> | Its using a KNOPPIX base right now, and I know that's Xenised |
| 18:57 | <murb> | setuid: send them a ready prephared image. |
| 18:58 | <murb> | if ness on a CD> |
| 18:58 | <murb> | btw you get much better perforance if you don't use file:// backed Xen instances. |
| 18:58 | <setuid> | Here's screenshots of what I'm doing (took these yesterday) |
| 18:58 | <setuid> | http://photos.gnu-designs.com/v/debuntanglexen/ |
| 18:59 | <setuid> | The guest vm takes out all networking to the box after this step: http://photos.gnu-designs.com/v/debuntanglexen/untangle_setup9.png.html |
| 19:00 | |-| | rharper [~rharper@bi01p1.co.us.ibm.com] has quit [Quit: Leaving] |
| 19:01 | <setuid> | "Saving Final Configuration", bam. Host networking resets. |
| 19:03 | <setuid> | whoa |
| 19:03 | <setuid> | I got it |
| 19:04 | <setuid> | er... well, it installed, but no networking inside the vm |
| 19:05 | <murb> | silly question, why not run this tangled thing on the raw metal? |
| 19:07 | <setuid> | No idea, I was tasked with getting it working virtualized, and since I've done this thousands of times with other vt products, I figured Xen couldn't be too difficult |
| 19:08 | <murb> | maybe they gave you this take because they failed themselves? |
| 19:08 | <setuid> | probably |
| 19:08 | <murb> | setuid: but best bet would be to install the newer debian packages from lenny |
| 19:08 | <murb> | so you've got a xen 3.1 base |
| 19:09 | <murb> | then try again with the hvm. |
| 19:09 | <murb> | it isn't your problem if it is bloody slow etc. |
| 19:09 | <murb> | except it eats your time testing it. |
| 19:09 | <setuid> | I've got it working in 3.0.3, had to remove the ioemu line from the vif line. I've shut down the vm, and will put those back, and see if it works |
| 19:10 | <murb> | so it was the vif line that scewed it up? |
| 19:10 | <setuid> | Well, the vif line somehow affected the host's networking, which was somehow triggered in the guest, knocking it out |
| 19:10 | <setuid> | I'm booting a fully installed-and-configured guest now, with the original vif line, will see what it does |
| 19:12 | <setuid> | Ok, putting back the ioemu bits, booting the guest, I still have networking... so something in the vif line during setup, was knocking it ou |
| 19:13 | <murb> | setuid: having both interfaces on the same bridge looked a bit suspect. |
| 19:14 | <setuid> | But it works after install? |
| 19:14 | <murb> | not a normal thing you'd execpt so you might have hit a corner case in the scripts. |
| 19:16 | [~] | setuid punches holes through the fw to get this outside for testing |
| 19:17 | |-| | hbaum [~hbaum@bi01p1.co.us.ibm.com] has quit [Quit: Client exiting] |
| 20:15 | |-| | ahs3 [~ahs3@66.37.132.58] has quit [Quit: off in the ozone, again] |
| 20:16 | <setuid> | Question: How hard is it to get vnc working to the guest vm, vs. using sdl-over-X11? |
| 20:38 | <setuid> | I just set vnc=1, vncviewer=1, sdl=0 in the config, and restarted the guest, but I can't get xvncviewer to attach to it |
| 20:38 | <setuid> | Is there some magical incantation? |
| 20:41 | |-| | zul [~chuck@CPE0006258ec6c1-CM000a73655d0e.cpe.net.cable.rogers.com] has quit [Remote host closed the connection] |
| 20:47 | <icarus901> | setuid, time to set vnclisten='0.0.0.0' OR keep it at default which is localhost and ssh -L port:127blah:port user@host; vncviewer localhost:port |
| 20:47 | <setuid> | uh |
| 20:47 | <setuid> | No, I mean I want remote clients to be able to attach to the running graphical vm, using xvncviewer on the client end |
| 20:47 | <icarus901> | i understood |
| 20:47 | <setuid> | No ssh involved, no wacky port forwarding involved |
| 20:47 | <icarus901> | this is what im talking about |
| 20:48 | <icarus901> | first one makes the port listen on all interfaces |
| 20:48 | <icarus901> | second option requires tunneling (if you keep the default vnclisten value of 127.0.0.1) |
| 20:49 | <icarus901> | so set it in the config (or in /etc/xen/xend-config.sxp to vnclisten='0.0.0.0' or the ip of the interface, etc |
| 20:49 | <setuid> | Ok, I think I got it working now |
| 20:49 | <icarus901> | then it will be available |
| 20:49 | <icarus901> | setuid, what the hell was the issue in the end |
| 20:49 | <icarus901> | (im still reading the above) |
| 20:51 | <setuid> | icarus901, It appears that having ioemu in the vif line during install, caused the guest to trigger a bus reset in the host. I removed the ioemu part, and the install completed, shut down the guest, put the ioemu back, and now it works |
| 20:51 | <icarus901> | oh how pleasant |
| 20:51 | <icarus901> | thats ridiculous |
| 20:51 | <setuid> | I agree, but that's all I did, and I used the core 3.0.3 packages from Debian |
| 20:52 | <setuid> | No 3.0.4 or 3.1.0 trickery |
| 20:52 | <setuid> | All stock packages from Debian Stable repos |
| 20:52 | <setuid> | Installed with this: |
| 20:52 | <setuid> | vif = [ 'bridge=xenbr1', 'bridge=xenbr1' ] |
| 20:52 | <setuid> | After install, changed to this: |
| 20:52 | <setuid> | vif = [ 'type=ioemu, bridge=xenbr1', 'type=ioemu, bridge=xenbr1' ] |
| 20:52 | <setuid> | And it seems to be working |
| 20:52 | [~] | icarus901 mentioned 3.0.4+ because it instantly fixed all HVM installs for him |
| 20:53 | <icarus901> | but that was a long time before debian included xen in their repos, much less a 3.0.3 with fixes backported from newer versions |
| 20:53 | <icarus901> | have you found any documentation or bugs or additional anecdotal evidence of similar issues? |
| 20:54 | <icarus901> | ive read the mailinglist for over a year and a half and i havent ever seen such a problem, unless i missed it |
| 20:54 | <setuid> | No, and I've been looking |
| 20:57 | <icarus901> | setuid, i'm really quite sorry i was unable to help |
| 20:57 | <setuid> | No no, you were very helpful... that and my blockheaded nature to not let something beat me ;) |
| 20:57 | <setuid> | I was about to give it up a day ago, and it turned out to be 1 line in the config. |
| 20:57 | <icarus901> | frankly i'm glad you didnt give up on it |
| 20:58 | <setuid> | Yeah, I'm like that. It's probably why I've been using Linux this long... I'm a masochist ;) |
| 20:58 | <icarus901> | you might want to attempt converting to a paravirtual machine for greater performance |
| 20:58 | <setuid> | Actually, I like to see things to closure, whether its flat-out broken, bugs identified, or some non-standard config fully tested. |
| 20:59 | <setuid> | Wait, hvm is slower than pv? |
| 20:59 | <icarus901> | yes it is |
| 20:59 | <setuid> | I thought "fully virtual" (using the hardware vt), would be faster than pv |
| 20:59 | <setuid> | Is the conversion easy? |
| 20:59 | <icarus901> | but in your case it was *necessary* to install an HVM machine, lacking any elegant way to install the domU natively paravirtual |
| 21:00 | <setuid> | Oh wait, so I'd need to somehow get a kernel recompiled in the guest, which is xen-aware, right? |
| 21:00 | <icarus901> | nope |
| 21:00 | <icarus901> | basically what you need to do is rip out the partitions into lvm or files, or disk, whatever you wish |
| 21:00 | <icarus901> | give it a generic paravirtual config |
| 21:01 | <icarus901> | point the config at the paravirt -xen kernel included with the debian packages |
| 21:01 | <icarus901> | modify a few things in the guest's /etc |
| 21:01 | <icarus901> | i'm not suer what the debian packages include, so you may need to futz with x |
| 21:02 | <icarus901> | the alternative is giving the hvm guest paravirtual drivers |
| 21:02 | |-| | rharper [~rharper@r74-193-70-77.pfvlcmta01.grtntx.tl.dh.suddenlink.net] has joined #xen |
| 21:02 | <icarus901> | to access the xen backend to skip modifying the rest |
| 21:02 | <setuid> | And how would I do that? (no paravirt-xen anything in any package in Debian, upstream or installed) |
| 21:03 | <icarus901> | the same kernel that you have running in dom0 right now will doe just fine |
| 21:03 | <setuid> | So in my guest's config, point to the same -xen kernel I"m using in dom0? |
| 21:03 | <setuid> | I guess I need to mount the guest disk.img file as a loopback, and rsync it off to some lvm partition on the host? |
| 21:04 | <icarus901> | if thats the way you wish to set it up, sure |
| 21:04 | <icarus901> | there are many permutations |
| 21:05 | <icarus901> | what bitness does the HVM domU have for userspace and kernel? |
| 21:05 | <setuid> | Well, a.) I don't know how to use lvm (though I rebuilt the host from scratch as an lvm system), b.) the xen guest disk image can't be mounted as a loopback apparently, c.) No idea what needs to be tweaked inside the guest vm to enable pv for that guest. |
| 21:06 | <icarus901> | it cant be mounted loopback because its a full disk |
| 21:06 | <icarus901> | with a bootsector and probably a few partitions |
| 21:06 | <setuid> | domU is some 2.6 kernel variant based on KNOPPIX, probably 32-bit of some ilk |
| 21:06 | <icarus901> | you could use kpartx from multipath-tools to pull out the partitions to loopX as block devices |
| 21:07 | <setuid> | kpartx sounds like a graphical tool... the box has no gui |
| 21:07 | <icarus901> | its not graphical |
| 21:07 | <setuid> | The host, that is |
| 21:07 | <icarus901> | i have no graphics on my host machine |
| 21:07 | <icarus901> | i ask about bitness because if youre using a paravirt 64bit kernel, the 32bit userspace for iptables etc wont like that very much |
| 21:07 | <setuid> | I fall into 'starts-with-k-must-be-kde' mode ;) |
| 21:07 | <icarus901> | i'm a console guy |
| 21:07 | <icarus901> | so assume its all cli |
| 21:08 | <setuid> | The end-game for this specific appliance, is to make sure that the host routes all traffic through the guest for filtering |
| 21:09 | <icarus901> | i understand |
| 21:09 | <icarus901> | and so you want to maximize performance |
| 21:09 | <setuid> | Right |
| 21:09 | <icarus901> | you could leave it as HVM |
| 21:09 | <icarus901> | it wont be lightning fast |
| 21:09 | <icarus901> | is 10mb sufficient? |
| 21:09 | <icarus901> | for the outbound pipe? |
| 21:09 | <setuid> | No, it sounds like pv is the way to go... just need to figure out how to convert it all over |
| 21:10 | <setuid> | Oh sure, 10M is fine |
| 21:10 | <icarus901> | then hvm might not be a problem |
| 21:10 | <setuid> | Well, let's call the hvm -> pv an "educating" experience then... because I can forsee a client asking for max performance (and I'd want to justify that to them anyway at some point) |
| 21:11 | <icarus901> | this will be a *very* educational experience then |
| 21:11 | <setuid> | hahah |
| 21:11 | <setuid> | Is there a howto out there? I'm happy to just sit and read and try things |
| 21:11 | <icarus901> | nope |
| 21:11 | <icarus901> | welcome to xen documentation hell |
| 21:11 | <setuid> | hehehe |
| 21:12 | <setuid> | Ok, well... I'm happy to write doco, because I do that anyway, as part of my other projects (and I'm documenting the hell out of what I've already learned through this experience) |
| 21:12 | <icarus901> | if i werent an hour away, id really prefer to do this over coffee haha |
| 21:12 | <icarus901> | anyway |
| 21:12 | <setuid> | Wait, whre are you? |
| 21:13 | <icarus901> | documentation would be fantastic actually, because (i may have mentioned) i've just started a xen documentation push with a few others |
| 21:13 | <icarus901> | storrs |
| 21:13 | <setuid> | ah right |
|
