Things are just odd. I have two identical ethernet ports on the motherboard. It boots from one (pxelinux.0). SUSE install kernel starts. Then, when it wants to access the install server, it does another dhcp request - but on the other port than it booted. Why, I do not know. I see that the install kernel only enabled one ethernet port. So I only have eth0. Anyone can guess which port gets selected. Selecting eth0 or eth1 on the kernel command line is pointless an only one (not the one it booted from) gets enabled (eth0). I would think that if one MAC address got an address, the dhcp server would not just give the address to the next different MAC address that comes along. Here is what I see in my logs: When booting via PXE: dhcpd: DHCPDISCOVER from 00:30:48:56:38:b3 via eth0 dhcpd: DHCPOFFER on 192.30.105.204 to 00:30:48:56:38:b3 via eth0 dhcpd: DHCPREQUEST for 192.30.105.204 (192.30.105.201) from 00:30:48:56:38:b3 via eth0 dhcpd: DHCPACK on 192.30.105.204 to 00:30:48:56:38:b3 via eth0 When SUSE install kernel does a DHCP request just before locating the install server: dhcpd: DHCPDISCOVER from 00:30:48:56:38:b2 via eth0 dhcpd: DHCPOFFER on 192.30.105.204 to 00:30:48:56:38:b2 via eth0 dhcpd: DHCPDISCOVER from 00:30:48:56:38:b2 via eth0 dhcpd: DHCPOFFER on 192.30.105.204 to 00:30:48:56:38:b2 via eth0 Note that the second group of commands is from a different ethernet port (?!?) and that the client never does a DHCPREQUEST after the server's DHCPOFFER I am a bit lost here... On Wed, 2006-02-22 at 18:56 +0100, Roger Oberholtzer wrote:
On Wed, 2006-02-22 at 18:40 +0100, Gunreben, Peter (Peter) wrote:
Roger,
Roger Oberholtzer wrote:
Roger,
Roger Oberholtzer wrote:
The install kernel boots. At some point it wants to get a
DHCP address
for the ethernet card. This 'should' be working as it got an address when it first booted. I see this in the server's /var/log/messages. But this seems to timeout. My dhcp entry is:
host jboc1 {
filename "/suseInstall/pxelinux.0"; fixed-address 192.30.105.204; hardware ethernet 00:30:48:56:38:B2; next-server 192.30.105.201; option host-name "jboc1"; }
192.30.105.201 is the server for everything (dhcp, pxe, suse install)
After this, I get a message that the SUSE installation source cannot be found. When trying to set it up by hand (nice it does that), I see that the default values it offers are the ones I expect (despite the probable DHCP oddness). But when it tries to mount the nfs install
On Wed, 2006-02-22 at 17:51 +0100, Gunreben, Peter (Peter) wrote: source, it
fails with 'unknown error -1'.
Have you checked whether the nfs-export works properly, e.g. by mounting it manually? I had a similar problem, where I tried to export a loop-mounted directory tree. Only, a userspace-nfsserver can do that.
I can mount it from other systems. Using
mount -t nfs source:/vol1/distro/OpenSUSE10.0/suse10.0 /tmp/vv
I can read files and go into directories.
It just got back to my mind that I had an additional problem. For any reason I still do not know, I had to add insmod=<my_NIC_kernelmodule> to the kernel parameters in order to get everything work. Of course, you need to replace <my_NIC_kernelmodule> with the one you need. May be it's worth to give it a try.
I have this: insmod=e1000
When all eventually fails, I see that the module is loaded. There seems to be activity on eth0 as well.
Is there no way to log access attempts for this? Beyond the dhcp stuff, which indicates, me thinks, that all is well.
The computer does have two identical ethernet ports. I have tried one or the other, as well as both. I could have made a mistake in doing that test, but I don't know how. I do not see a dhcp request on an unexpected port. Only on the one I expect.
I want to install 10 systems in the morning. And more after that. I didn't expect this part to be the problem. I was saving that for autoyast.
Peter.
-- Roger Oberholtzer OPQ Systems AB Ramböll Sverige AB Kapellgränd 7 P.O. Box 4205 SE-102 65 Stockholm, Sweden Tel: Int +46 8-615 60 53 Fax: Int +46 8-31 42 23