| ErKa (keryell@wireless-tcc-users-915.sc11.org) joined #fai. |
| ErKa (keryell@wireless-tcc-users-915.sc11.org) left irc: Read error: Operation timed out |
| ErKa (keryell@wireless-tcc-users-915.sc11.org) joined #fai. |
| ErKa (keryell@wireless-tcc-users-915.sc11.org) left irc: Ping timeout: 480 seconds |
| jonas (~jonas@115.78.133.75) joined #fai. |
| Nick change: jonas -> Guest17233 |
| Nick change: Guest17233 -> jonas |
| sneumann__ (~sneumann@pD9554DE6.dip.t-dialin.net) joined #fai. |
| sneumann__ (~sneumann@pD9554DE6.dip.t-dialin.net) left irc: Ping timeout: 480 seconds |
| sneumann__ (~sneumann@router-sub.ipb-halle.de) joined #fai. |
| ErKa (~keryell@71-37-54-135.tukw.qwest.net) joined #fai. |
| ErKa (~keryell@71-37-54-135.tukw.qwest.net) left irc: Ping timeout: 480 seconds |
| and1bm (~andi@HSI-KBW-095-208-066-227.hsi5.kabel-badenwuerttemberg.de) joined #fai. |
| shod (~shod@ip-109-47-246-97.web.vodafone.de) joined #fai. |
| Mrfai (~lange@malzmuehle.informatik.uni-koeln.de) joined #fai. |
| verwilst (~verwilst@d5152FEFB.static.telenet.be) joined #fai. |
| Mrfai (~lange@malzmuehle.informatik.uni-koeln.de) left irc: Read error: Operation timed out |
| Mrfai (~lange@malzmuehle.informatik.uni-koeln.de) joined #fai. |
| shod (~shod@ip-109-47-246-97.web.vodafone.de) left irc: Quit: Leaving. |
| shod (~shod@ip-109-47-246-97.web.vodafone.de) joined #fai. |
| shod (~shod@ip-109-47-246-97.web.vodafone.de) left irc: Ping timeout: 480 seconds |
| and1bm (~andi@HSI-KBW-095-208-066-227.hsi5.kabel-badenwuerttemberg.de) left irc: Quit: Konversation terminated! |
| carsten (~carsten@ahgate1.aei.uni-hannover.de) joined #fai. |
| 13:06 carsten | Hi all,as just written to the fai main list, has there been some work on getting multiple NICs after PXE booting under control with IPAPPPEND 3? |
| 13:13 carsten | at least live-boot-initramfs-tools 2.0.15-1 does not seem to address this and also the latest one from Wheezy (3.0...) not |
| 13:14 Mrfai | carsten: I just send an email about this. And mika is the expert for initramfs-tools and live-boot. |
| 13:16 carsten | i Thomas, just read it |
| 13:16 carsten | will look into that, would be nice to get this fixed |
| 13:16 carsten | it's becoming really annoying to us as we get more andmore different computers |
| 13:18 Mrfai | We did not have that many problems with multiple NICs here (in IRC) in the past months. So I guess, it should be solvable. |
| 13:18 Mrfai | But I'm sure mika will help you if there are still problems. |
| 13:19 Mrfai | s/mika/mikap/ |
| 13:19 mikap | well, i'm aware of this issue and know that people have different solutions solving this issue |
| 13:19 mikap | ours[tm] is that we're supporting ip=... syntax as used by the kernel/nfsroot and we don't have "rotating" NICs |
| 13:20 mikap | so the ones with working solutions should provide us (initramfs-tools/live-boot) their patches |
| 13:21 mikap | or instead provide access to such boxes, sadly i don't have machines with such problems available those days |
| 13:22 carsten | mikap: so far that worked well for us |
| 13:22 mikap | so far as until what? ;) |
| 13:23 Mrfai | IMO the important device name during installation is only eth0, which should be the NIC, which received the kernel and initrd. |
| 13:23 carsten | mikap: however, now we have machines which uses a different eth0 after each reboot |
| 13:23 mikap | carsten: what vendor is that? |
| 13:23 carsten | supermicro |
| 13:23 mikap | tk? ;o) |
| 13:24 carsten | X8DT6 board with 2 82598EB 10Gb/s NICs and 2 82574L 1 b/s NICs |
| 13:24 Mrfai | carsten: there's a udev rule that should care of stable NIC names IIRC. /etc/udev/rules.d/70-persistent-net.rules |
| 13:24 mikap | Mrfai: actually that's not necessarily true, especially if booting environment is different from the one providing the deployment data (e.g. special vlan) - but that's at least what should work always ootb, yeah |
| 13:24 carsten | yes, but udev does not run in the initd before mounting the nfsroot, right? |
| 13:25 carsten | the only problem we really face is that PXE works, but the initrd tries the wrong NIC for mounting the nfsroot (DHCP fails/SHCP stalls) |
| 13:26 carsten | ok, brb, meeting is over and I need to move back to my office |
| 13:26 carsten | .quit |
| carsten (~carsten@ahgate1.aei.uni-hannover.de) left irc: Quit: Konversation terminated! |
| carsten (~carsten@ahgate1.aei.uni-hannover.de) joined #fai. |
| 13:30 carsten | back again |
| 13:34 carsten | Mrfai: thus I don't think udev might help much here, right? |
| 13:37 Mrfai | I've looked at the script/function code. If BOOTIF is set, the initrd can detect the device name from which the kernel and initrd was loaded. So, it does not need to be eth0. It uses the MAC address to get the device name of the NIC for later nfs mount. |
| 13:38 zeha | (Note that the grml initramfs has different code.) |
| 13:39 Mrfai | Then, the installation should generate the 70persistent-net.rules file, which should care about stable device names during the next reboot. That's how I understand these things. |
| 13:41 Mrfai | zeha: Mmm, initramfs-tools_0.93.4-grml02 has no BOOTIF code in it :-( |
| 13:42 mikap | Mrfai: it's not about initramfs-tools but live-initramfs/live-boot |
| 13:43 Mrfai | mikap: I don't think so. If initramfs-tools is setting DEVICE to the correct interface which received the kernel, then live-boot can just use this device name for booting the nfsroot. |
| 13:44 mikap | Mrfai: the problem is that live-initramfs upstream re-invented networking code instead of reusing what's there |
| 13:44 Mrfai | This is more flexible, than that the admin must specify a device name (which might change) |
| 13:45 zeha | well, there's a lot of stuff going on between live-i-t and i-t, you'll have to debug it to follow the code flow |
| 13:45 mikap | yes, that's the IPAPPEND way which is fine, though you *might* want to apply specific configuration statically to a device, that's when ip=... is nice then |
| 13:45 carsten | Exactly, if IPAPPEND 3 is given via PXE and the kernel and initrd is loaded, I *think* the sane way is to re-use this for bringing that iface up and use for nfs mouting |
| 13:46 zeha | the code does ip= and BOOTIF=; maybe not both at the same time |
| 13:48 Mrfai | carsten: did you try those packages in your nfsroot? initramfs-tools 0.98.8, live-boot-initramfs-tools, live-boot 2.0.15-1 |
| 13:49 Mrfai | I'm using thoses for squeeze. They include the BOOTIF code |
| 13:49 carsten | the latter tow only |
| 13:49 carsten | initramfs-tools is NOT installed |
| 13:49 carsten | wait, sorry |
| 13:49 carsten | ii initramfs-tools 0.98.8 |
| 13:49 carsten | yeah all three of these |
| 13:50 carsten | jsut started a test install and initrd is hanging again with this line in the pxeconfig line |
| 13:50 carsten | append initrd=initrd.img-2.6.32-5-amd64 root=/dev/nfs nfsroot=/srv/fai/nfsroot boot=live console=tty0 nomodeset FAI_FLAGS=verbose,sshd,createvt,initial,reboot FAI_ACTION=install |
| 13:50 carsten | IPAPPEND 3 |
| 13:51 zeha | make sure your klibc-utils package is version 1.5.20-1+squeeze1 |
| 13:52 zeha | older ipconfig versions will definitively cause issues with >1 nic |
| 13:52 carsten | ok, that is only at 1.5.20-1 |
| 13:53 carsten | suggestions: re-create NFSroot or safe-upgrade in chroot? |
| 13:54 Mrfai | recraete nfsroot, since then a new initrd is also created and copiesd to the tftp directory. You may want to use m-f-n -k or -K |
| 13:55 carsten | ok, that might take a while (like Thomas is warning ;)) |
| 14:12 Mrfai | carsten: Wenn das alles nicht klappt, koennte dracut oder ein squashfs (gebaut mit fai-nfsroot2image) helfen. |
| ranulf (~ranulf@mail3.twallrafen.de) joined #fai. |
| 14:19 carsten | Mrfai: ok, wait and see, nfsroot ready, reboot of test node |
| 14:22 carsten | hmm, same, stops suring select-eth |
| 14:22 carsten | during |
| 14:22 carsten | Let me just look at the initrd if I can spot bootif in there |
| 14:24 Mrfai | /quit/quit |
| Mrfai (~lange@malzmuehle.informatik.uni-koeln.de) left irc: Quit: leaving |
| 14:25 carsten | darn, I scared him away ;) |
| 14:26 oz_ | hehe |
| 15:01 carsten | ok, lenghty email to the main fai list |
| 15:02 carsten | it seems the function configure_networking from scripts/functions is never called (at least not in my case) and thus BOOTIF is never evaluated |
| 15:02 carsten | :( |
| 15:17 zeha | probably. |
| 15:18 zeha | if you have boot=live then you're after scripts/live networking code |
| 15:34 carsten | yeah, thus Thomas was wrong :( (and without boot=live it falls into the initramfs shell) |
| 15:35 carsten | *sigh* |
| 16:01 carsten | mikap: I'm just wondering is there a deeper sense in having "export BOOTIF=" in "init" within the initrd? this effectively disables this feature, right? |
| 16:03 carsten | mikap: at least it heavily confuses me - I added 'env' to 'init' and before the export the variable is set corerctly and afterwards it's empty |
| verwilst (~verwilst@d5152FEFB.static.telenet.be) left irc: Quit: Ex-Chat |
| carsten (~carsten@ahgate1.aei.uni-hannover.de) left irc: Quit: ... |
| jmux (~jan-marek@lhm246.muenchen.de) joined #fai. |
| sneumann__ (~sneumann@router-sub.ipb-halle.de) left irc: Read error: Operation timed out |
| and1bm (~andi@HSI-KBW-095-208-066-227.hsi5.kabel-badenwuerttemberg.de) joined #fai. |
| ErKa (keryell@wireless-tcc-users-915.sc11.org) joined #fai. |
| jmux (~jan-marek@lhm246.muenchen.de) left irc: |
| ErKa (keryell@wireless-tcc-users-915.sc11.org) left irc: Ping timeout: 480 seconds |
| jonas (~jonas@115.78.133.75) left irc: Read error: Connection reset by peer |
| and1bm (~andi@HSI-KBW-095-208-066-227.hsi5.kabel-badenwuerttemberg.de) left irc: Quit: Konversation terminated! |
| sneumann__ (~sneumann@pD9554DE6.dip.t-dialin.net) joined #fai. |
| ErKa (keryell@wireless-tcc-users-915.sc11.org) joined #fai. |
| ranulf (~ranulf@mail3.twallrafen.de) left irc: Quit: leaving |
| sneumann__ (~sneumann@pD9554DE6.dip.t-dialin.net) left irc: Ping timeout: 480 seconds |
| ErKa (keryell@wireless-tcc-users-915.sc11.org) left irc: Ping timeout: 480 seconds |
| --- Thu Nov 17 2011 |