PXE/TFTP'ing 9.1 on AMD64
Hello, I currently have SuSE 9.1 Pro PXE/TFTP booting/autoinstalling working great with our 32-bit machines. My problem is that I can't seem to get this to also work with our x86_64 machines. I don't know if there is a graceful way to separate 32 and 64-bit stuff in the /tftpboot directory, so what I initially tried to do was make an isolinux.cfg menu option which specified an alternate (64-bit) linux and initrd, but when it loads the 64-bit kernel and initrd, the machine immediately reboots. I tried to boot it off the 32-bit kernel/initrd just to do the install, but as soon as it loads the data into ramdisk, it goes to a blue yast screen with the message "An error occurred during installation!" (very helpful) Lastly, I backed up the /tftpboot directory and blew it away, making a brand new one from scratch with the 64-bit distribution, and it behaved exactly like it did when I specified the alternate linux and initrd (rebooted). Now this situation suggests that setting up 64bit network booting the same way as with the 32-bit stuff doesn't work with my 64-bit systems, and that there is something special I need to be doing for them. Is there some documentation specific to 64-bit network booting / autoinstalling? Does it in fact differ from the 32-bit procedure (aside from having an x86_64 kernel/initrd/distro/control-files)? Has anyone successfully PXE/DHCP/TFTP/autoinstalled an AMD64 machine? Perhaps it is just a compatibility problem with my hardware? The 64bit systems I'm installing are Opteron 146's on Asus SK8N motherboards (nForce chipset). Thanks in advance, -Charles Smith
On 29 Jun, Charles Smith wrote:
I don't know if there is a graceful way to separate 32 and 64-bit stuff in the /tftpboot directory, so what I initially tried to do was make an isolinux.cfg menu option which specified an alternate (64-bit) linux and initrd, but when it loads the 64-bit kernel and initrd, the machine immediately reboots.
Just use different filenames... the pxe/tftp steps just read what you say. I use host specific config files, because our auto-install system uses a name=<hostname> arg to get the name and setup the last host stuff, but a sample config is: #DEFAULT qct-64bit name=rs-workstation #DEFAULT qcti name=rs-workstationDEFAULT disk DEFAULT disk LABEL qcti KERNEL vmlinuz APPEND nofb ks=nfs:192.168.1.1:/local/mnt/RedHat-7.3/GV/ks/ks-qct2x36i.cfg initrd=initrd.img ksdevice=eth1 network LABEL qct-64bit KERNEL vmlinuz-sp3 APPEND load_ramdisk=1 initrd=initrd-sp3 autoyast=nfs://192.168.1.1/local/mnt/x86_64/sles8/GV/ks/rs-blade.xml netdevice=eth1 insmod=bcm5700 install=nfs://192.168.1.1/local/mnt/x86_64/sles8 LABEL disk LOCALBOOT 0 In this example, qcti is a 32bit install, qct-64bit is 64bit.. and disk is boot off disk. I use this setup for blades in our blade racks. All our blades pxe boot every time. When installing, the DEFAULT disk line is commented out and the appropriate image is not. During the install process, I have a post(or pre) install script mount the /tftpboot path off the server, and then change it's pxeboot config file to comment out the install image, and uncomment the disk call.. then on future boots, it boots off local disk. But if I want to re-install the OS, I can just edit the config file and reboot the host. But as you can see from above, the 32bit image uses vmlinuz and initrd as the filename, and the 64bit one uses vmlinuz-sp3 and initrd-sp3 as it's file names. Allthough, now that I read your mail again, I see that you might've already done this and it just rebooted on you. You're not using arima motherboards are you? If so, what's your bios rev? We had some issues with older revs with more then 4gig of ram and memory bank interleaving turned on. -- Mike Marion-Unix SysAdmin/Staff Engineer-http://www.qualcomm.com Homer: "Bless you boys." [He's saluting the men marching by in the parade] Marge: "Homer, those are ice cream men." [Homer tears up] Homer: "I know." ==>Simpsons
Just use different filenames... the pxe/tftp steps just read what you say.
Yeah, that's what I'm doing. I have a -64 version of each that I specify in the boot menu like: label 64bit-install kernel vmlinuz-64 append initrd=initrd-64 ramdisk_size=65536 splash=silent showopts install=nfs://192.168.0.224:/suse/x86_64/current/ autoyast=nfs://192.168.0.224:/control-files/x86_64/ It seems like that's probably the best way to go, I just didn't know if there was a way to have parallel /tftpboot directories for different architectures. I never thought about all the scripting to change the default. I just always have mine default to "harddisk", but then again, we don't have many headless machines, either.
Allthough, now that I read your mail again, I see that you might've already done this and it just rebooted on you. You're not using arima motherboards are you? If so, what's your bios rev? We had some issues with older revs with more then 4gig of ram and memory bank interleaving turned on.
No, I'm using ASUS SK8N motherboards. The BIOSes were flashed from something-004 to something-007 (the latest). We actually just now tried playing with the memory settings in the BIOS (Turned off a lot of the ECC memory settings) and it finally started loading the kernel and initrd, and going through the whole install! But now when it reboots, it kernel panics when it tries to mount the reiserfs partitions. I'm not using the onboard Promise controller or anything, and all the BIOS hard drive settings are set to auto. Booting to S and running a manual reiserfsck makes it puke, complaining about the DMA settings.. I've got lots of things to try still, but at least I'm not having the original problem anymore. Thanks for your response! -Charles Smith
participants (2)
-
Charles Smith -
mmarion＠qualcomm.com