Hi Allan: When you reply, can you please keep the same subject line? This will allowing archive threading easier (also better for my gmail reader), thanks!
On 2/10/07, Allan Menezes <[EMAIL PROTECTED]> wrote:
I am using the command ./install_cluster eth1 for generation of the oscar wizard and install interface. Further more I have two gigabit switches per eth interface and all nodes have two gigabit eth interfaces.
A little bit of clarification. I understand that your headnode + all clients have at least 2 networks cards and you want them all connected via two switches. YES, That is right. ONly two compute nodes of the 14 have P5B-VM DO motherboards, the resthave asrock 775i65G motherboards with built on 10/100 eth that i disable in BIOS. I tried one of them as a client node and it booted and installed okay! These asrock motheboards have two nics one skge or r8169 and the other intel e1000
on two pci slots plugged in. The P5B-Vm DO mobo are new mobos from asus and have support in fc6 which is why i am anxious to use fc6. These asus mobos have built in 965 chipset with intel graphics controller and 6 channel sata-II raid controller. In the master node with one of these asus boards i have two 250gig sata-2 drives with raid 0 for 500gig total capacity. Both these mobos (asus p5b-vm do) have built on pci-express gigabit ethernet card and one e1000 pci32 pro100/gt intel card that i plugged in. I have two subnets IP's 192.168.3.x and 192.168.1.x and each subnet is connected to a netgear gigabit 24 port high speed switch for 14 nodes. eth1 is the pci32 addon intel card on the 192.168.1.x subnet. The prblem is that these asus mobos being new supported by the 2.6.18 kernel and onwards i may have to pass command line options on the boot: prompt such as: boot: kernel all-generic-ide pci=nommconf root=/dev/sda1 irqpoll acpi=off If i do pass boot prompts as above i get a kernel panic before the dhcp etc But if i do not pass a boot prompt as above it goes on nicely till it fails on the dhcpdiscover though the mac address of the eth interface is set up right in Setup Networking of the boot client and it seems to get its ip from pxe boot and the main server 192.168.1.1 I do not understand what's wrong! It's only the new mobos with dual intel nics. OSCAR can only help you set up with ONE interface, configuring and setting up the other interface will be up to you to set up. Although, I guess with some modifications to our code, it may be possible to help you set this up automatically. I diabled in bios the motherboard nic and i still get the same problem even disconnected the cable from the mobo!
The Fc6 i386 installed on the P5B-VM DO mobo has two intel interfaces with eth1 as ip 192.168.1.1 and a pci nic installed on a pci slot. eth0 is an onboard pci express nic eth interface with ip 192.168.3.1. I had found the cables switched on the two switches( individual switches per interface) but I corrected that and rebooted both machines and have the same problem.
Since you ran ./install_cluster eth1, eth1 will be the only interface which OSCAR cares about. So you need to plug the first interface (eth0) on your P5B-VM DO computer onto the switch that your headnode's eth1 is plugged into, and ensure that the MAC address of your client computer's eth0 is recorded in OSCAR's "Setup Networking" widget. I don't see why dhcp wouldn't work in this case. BTW, if you have access to an IRC client, perhaps this discussion is better done interactively. You can find me at #oscar-cluster on irc.freenode.net. Cheers, Bernard
------------------------------------------------------------------------- Using Tomcat but need to do more? Need to support web services, security? Get stuff done quickly with pre-integrated technology to make your job easier. Download IBM WebSphere Application Server v.1.0.1 based on Apache Geronimo http://sel.as-us.falkag.net/sel?cmd=lnk&kid=120709&bid=263057&dat=121642
_______________________________________________ Oscar-devel mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/oscar-devel
