On Thu, 2002-09-26 at 18:15, Markus.Loeffler@... wrote:
> Hello again ,
> first, thank you for the immediate help. I'm trying to get a SSI Cluster
> running to
> scale a system for a customer and i have some time-problem .......
> sorry for the bad english too ...
That's great !!!.=20
> i didn't find any netmask field in the /etc/clustertab. But copying the
> kernel to
> /tftpboot and rebuilding the ramdisk did the job.
True netmask is identified when you build the ramdisk. It is inside
the ramdisk (/initrd/etc/boottab ) That's why replacing the tftp image=20
> But i think i have some general Problems here, mainly in understanding th=
> whole thing :
> I configured my first Cluster node with RH72. Installed the Vanilla Kerne=
> applied the 0.7.1 patch.
I haven't run the SSI cluster on RH7.2 after rc.sysinit.nodeup was
introduced.I do most of my SSI work on debian But i will try to
answer some of the questions.=20
> Compiled the Kernel as in the SSI Install Instructions.
> Then in did the make install_ssi_redhat and configured the first node.
> Edited /etc/fstab and
> replaced LABEL=3D/ with my real device name of the root device
> (in my case /dev/cciss/c0d0p4, Compaq Smart Array Controller)
> Then edited /etc/xinetd.d/tftp, disabled kudzu, ran cluster_mkinitrd,
> edited lilo.conf and ran
> lilo-c -> successful. The mountpoint /devfs was alread there (?!).
> Again edited /etc/fstab to mount ssidevfs. Commented out (with #) the
> /dev/pts line.
> Here starts the confusion :
> generated the network config for the first node
> (cp -a /etc/sysconfig/network-scripts /etc/sysconfig/network-scripts-1)
This is how the SSI right now handles the context dependent symbolic
links. ( CDSL ). CDSL's are done by using the --bind option of the mount
command. But i guess this is going to change. . Now on each node
/etc/sysconfig/network-scripts should point to
> Do i understand this right, this is the Network-config for my first,
> already running node ?
> So i don't have to change anything in the network-scripts-1 script ?
yes see above.=20
> Then i made the entries in /etc/rc.d/rc.sysinit
> again :
> Do i understand this right, the additional nodes get their nodenumber via
> clusternode_num and
> should execute the script network-scripts-`clusternode_num` ?
NO rc.sysint is run only in the case of master node.Rest of the nodes
And if you see mounting the network-scripts is commented . Uncomment
> Now i generated the network config for the second node
> (cp -a /etc/sysconfig/network-scripts /etc/sysconfig/network-scripts-2)
> edited the eth0-config with the new ip address for the second node.
> Do the addnode-thing, nodenumber 2, IP Adress like in the eth0-config fil=
> When i start my second node (with bootdisk generated from rom-o-matic.net=
> it starts,
> but gets the same hostname as my first Node (clnode1) is this correct ?
> ifconfig shows that this node has the ip address i entered in addnode. so
> far so good.
> when i look at "df -ah" i see that the second node has mounted
> network-scripts-1 (not network-scripts-2 as i expected)
> is this correct ?
> I configure the swap device on the second node exactly with the same
> devicename as on the first node.
> Then i configure the boot device, fdisk and mkfs.
Swap is not getting mounted by default in the case of dependent node.
you will have to add in manually.This is going to change soon. I have
already checked in the code to make the fstab cluster wide.=20
> I run chnode on the first node, change the boot device to the new
> configured (dev/cciss/c0d0p1), do
> not make the second node to potential CLMS Master.
> This fails with "Error migrating to node 2. Try again."
I didn't understand this part. ? What is that you are trying to do ?=20
> Whats wrong ?
> ....dazed and confused ....
> Mit freundlichen Gr=FCssen
> Markus L=F6ffler
> - Systemengineer -