I'm upgrading my cluster from Rocks 4.2.1 to 5.0, and stuck
at the last step of frontend installation. The screen shows
that all packages have been installed, and it has been
"Running post-install scripts" for over an hour.
I selected all rolls except sge, bio and xen in the DVD, plus
my "Restore roll", Intel roll from ClusterCorp and Torque
roll.
What's wrong with my installation?
Thanks!
Huiqun Zhou
hit <Ctrl><Alt>F2 -- you should get a bash prompt.
then send us the output of:
# cat /tmp/ks-script
- gb
When I came to my office this morning, I found that the scripts have run
successfully to the end after they took almost 3 hours. It seems that
everything
is good except that I cannot find /boot/kickstart/cluster-kickstart-pxe for
re-
installing compute nodes as the manual indicated. But, there is a
/boot/kickstart/cluster-kickstart. Any ideas?
I think I can remove all my compute nodes and reinstall them again using
insert-ethers as it's small cluster with only 16 compute nodes. Are there
any problems here?
Huiqun
I got a problem when I started to reinstall my compute nodes by using
insert-ethers. My compute nodes seem to be failed in kickstarting, and
end with a screen requesting selection of language, keyboard ...
My compute nodes are DELL 1950 (dual core), they worked perfectly
with Rocks 4.2.1 ...
Are there anything I did wrong?
make sure the boot order on your compute nodes is:
CD
PXE
Hard DIsk
- gb
Huiqun
----- Original Message -----
From: "Greg Bruno" <greg....@gmail.com>
To: "Discussion of Rocks Clusters" <npaci-rocks...@sdsc.edu>
Sent: Wednesday, August 20, 2008 9:43 PM
Subject: Re: [Rocks-Discuss] Installation stuck
at"Runningpost-installscripts"
BTW, Greg, the problem I encountered with 4.2.1, when using
different basename but starting from same cabinet number, it could
mess up DNS, has been solved in 5.0. Thanks for the effort.