Re: [BProc] How do I make a beoboot kernel that will work with beoboot-cm.1.5?

[Larry B. <ba...@us...>]

Thanks to Josh Aune and Mike Madore for their help.

I have updated my Clustermatic 3 installation with the latest versions of
cmtools and beoboot.  I made notes for myself which I have included below.

My intent is to modify the new beoboot so it behaves more like it did on
Clustermatic 2.  In particular, I used to be able to NFS mount several file
systems -- shared read-only, shared read-write, and node-specific read-writ=
e
-- using a single /etc/beowulf/fstab:

    #
    # Read-only mount points
    #
    $MASTER:/bin    /bin            nfs     ro,nolock,rsize=3D8192    0 0
    $MASTER:/opt    /opt            nfs     ro,nolock,rsize=3D8192    0 0
    $MASTER:/sbin   /sbin           nfs     ro,nolock,rsize=3D8192    0 0
    $MASTER:/usr    /usr            nfs     ro,nolock,rsize=3D8192    0 0
    #
    # Private read-write mount points
    #
    $MASTER:/var/node.$NODE /var    nfs     rw,nolock,rsize=3D8192,wsize=3D8192
0 0
    #
    # Shared read-write mount points (MPICH 1.2.4, section 4.11.1: use
"noac")
    #
    $MASTER:/home  /home           nfs
rw,nolock,rsize=3D8192,wsize=3D8192,noac   0 0

This worked in Clustermatic 2 because setupfs was a shell script, and it
expanded each line in fstab before using it.  Clustermatic 3 changed that.
($MASTER has to be changed to MASTER; setupfs substitutes the master node's
IP address for MASTER for NFS mounts; no other substitutions are supported.=
)

I have tested code that defines the MASTER and NODE variables in the
Clustermatic 3 beoboot-cm.1.4 node_up program, and I wrote a beoboot plugin
that supports variable substitution while copying /etc/beowulf/fstab to the
node's /etc/fstab.  Next I will modify setupfs to merge that code into it
and then have setupfs work from the node's /etc/fstab.  I'm going to do tha=
t
work on the beoboot-cm.1.5 code, which I will publish when it is done.

I have never used the Clustermatic 2 setupfs feature that looks for
fstab.$NODE, but I have some ideas about restoring that feature in
Clustermatic 3.  I have to look at the node_up code in beoboot-cm.1.5 to se=
e
what has changed from beoboot-cm.1.4.  Also, it would be nice to expand
$MASTER and $NODE in node_up.conf, e.g., for node-specific miscfiles.

How important is support for $MASTER and $NODE expansion for fstab.$NODE or
in node_up.conf to you?


Larry Baker
US Geological Survey



LANL Clustermatic 3 Update


Install Clustermatic 3.  (See LANL Clustermatic 3 Install.)

Download the cmtools-1.2 and beoboot-cm.1.5 updates from
http://bproc.sourceforge.net to /usr/local/src.

Make an RPM for cmtools-1.2 and install it:

    # rpmbuild =ADtb /usr/local/src/cmtools-1.2.tar.gz
    # rpm =ADUvh /usr/src/redhat/RPMS/i386/cmtools*.rpm

Insert the LANL Clustermatic 3 CD in the CD reader and mount it, or mount
the downloaded ISO image directly:

    # mount =ADt iso9660 =ADr /dev/cdrom /mnt/cdrom

    or:

    # mount =ADt iso9660 =ADr =ADo loop /usr/local/src/CM3.iso /mnt/cdrom

Install the Clustermatic 3 kernel source RPM:

    # rpm =ADUvh /mnt/cdrom/SRPMS/kernel-2.4.19-lanl.22.src.rpm

Unmount the CD/ISO image file:

    # umount /mnt/cdrom

Enable tmpfs support in the beoboot phase 1 kernel:

    # cd /usr/src/redhat/SOURCES
    # vi kernel-2.4.19-i386-beoboot.config

    Change

        # CONFIG_TMPFS is not set
    to

        CONFIG_TMPFS=3Dy

Make the LANL 2.4.19 kernel RPMs and install the new beoboot phase1 kernel:

    # rpmbuild =ADba /usr/src/redhat/SPECS/kernel-2.4.spec
    # rpm =ADUvh =AD-force /usr/src/redhat/RPMS/i386/kernel-beoboot*.rpm

Make the beoboot-cm.1.5 RPMs and install them:

    # cd /usr/src
    # ln =ADs linux-2.4.19-lanl.22 linux
    # rpmbuild =ADtb /usr/local/src/beoboot-cm.1.5.tar.gz
    # rm linux
    # rpm =ADUvh /usr/src/redhat/RPMS/i386/beoboot*.rpm

Create a beoboot floppy disk for the phase 1 boot step for each slave node:

    # beoboot =AD1 =ADk /boot/vmlinuz-2.4.19-lanl.22beoboot =ADf =ADo /dev/fd0

Create the phase 2 boot images:

    # beoboot =AD2 -n

Reboot the master node and boot the slave nodes using the beoboot floppy
disks.

Verify the cluster configuration:

    # bpstat =ADal