Scali Manage – the bloated and unpredictable HPC cluster management tool – has recently been acquired by Platform Computing Corp. It is to be integrated into Platform LSF and Platform OCS products. I can’t say I am a big fan of Scali Manage. To begin with, I simply see no good reason for its existence. It is essentially a parasite OS sitting on top of Linux and sucking the life out of it.

If your system administration experience is limited to Windows point-and-click stuff, then maybe you will find Scali Manage handy. It has a GUI with tremendous functionality. It is also tremendously buggy. Scali’s main enemy is the system administrator. How else would you explain the fact that this application will attempt to reconfigure your machine on every reboot to undo any changes you’ve made?

Scali Manage does offer a few useful features: ability to push out OS images to cluster nodes, parallel command shell, easy way to connect to ILO ports of your cluster nodes, and a few other cool things. In my opinion all this extra functionality is not worth the adjustments you need to make to your sysadmin and your open source inventory management software configurations. Complaining aside, my customer bought a cluster that came with Scali preinstalled and I had no say in the matter. Below is a quick-and-dirty CLI command guide for Scali Manage that will allow you to bypass their buggy GUI for most common sysadmin tasks.

SCALI Cheat Sheet

Add NAT

scalimanage-cli help addnatservice

addnatservice [systemnames] [interface=eth1] Add NAT service to system(s)
Arguments:
systemnames – name of system(s) {[..]}
interface – name of the internal interface to NAT from.

Options:
–interface=INTERFACE

Add remote FS

scalimanage-cli help addremotefs

addremotefs [systemnames] [fstype] [src] [mntpoint] Add mounting for remote filesystem on system(s)
Arguments:
systemnames – name of system(s) {[..]}
fstype – type of filesystem, legal values: “nfs” “lustre”
src – source
mntpoint – mountpoint

Add NIS

scalimanage-cli addnisclientservice ln00320 easthartford “ln010 ln011.

The command line sequence is as follows:

scalimanage-cli createcluster [name of cluster]
scalimanage-cli addnodetocluster [systemnames] [clustername]
scalimanage-cli reconfigure all

Add a subnet

scalimanage-cli addsubnet [name] [subnetnumber] [netmask]

Cature and deploy images

scalimanage-cli captureimage [systemname] [imagename] [description] [excludes...]

where
systemname – string containing name of node to be imaged
imagename – what to call the image
description – description of an image; Default none
excludes -list of systems to be excluded

scalimanage-cli installimage [systemnames] [imagenames] [installserver]

where
systemname – name of system(s)
imagename – name of image
installserver – server from where the installation job will run.

Install & Capture Images

scalimanage-cli listimages

scalimanage-cli installimage [nodes] [imagename] [templatename]

To capture an image use:

scalimanage-cli capturelimage node [imagename]

Show Enabled Services

scalimanage-cli listhostedservices [name of head node]

Enabling RSH

scalimanage-cli addrshservice all

scalimanage-cli reconfigure all

Rename a Cluster

scalimanage-cli listclusters

And :

scalimanage-cli listnodes

You shouldn’t see identical names.

To rename a cluster, do :

scalimanage-cli renamecluster [oldclustername] [newclustername]

Name it something unique, for example if it has the same name as the head node, call it [headnode]-cluster

Then run /etc/init.d/scance restart on the head node, and finally the scalimanage-cli reconfigure all. command.

Enable/Terminate Console

console [nodename]

To terminate the console, press Ctrl-e c . (control e – c – dot ).

Clear Mac Address (HW replaced)

scalimanage-cli clearmacaddress rXnXX ethX

Get Image Template & Compare

scalimanage-cli gettemplate [UUID1] ]/tmp/foo” and “scalimanage-cli gettemplate [UUID2] ]/tmp/bar”

Add Nodes to a cluster and remove systems

You can expand the existing cluster, and you can also right click the added node and select “Delete system” to remove it from the configuration. Pretty easy.

You can also use CLI :

scalimanage-cli createnode [name] ...

scalimanage-cli addnodetocluster [name] [cluster]

scalimanage-cli removesystem [name]

PBS Status

IF PBS is configured through Scali Manage, we manage the /etc/pbs.conf file so it shouldn’t be edited manually. To verify that Scali Manage is managing the PBS configuration, please do the following:

scalimanage-cli listinstalledsoftware [headnode]

scalimanage-cli listhostedservices [headnode]

PBS to use SSH & SCP

PBS_SCP=/usr/bin/scp
PBS_RSHCOMMAND=ssh

PBS Quick Test

Submit a job through PBS and verify that it was executed successfully (As a regular user: scasub ls. Run qstat to see it listed in the queue and after a while it should go away and there should be two new files in the current directory: ls.XXo and ls.XX.e. The first one should include the file listing and the second should be empty.

Password Change

scalimanage-cli setrootpassword n[01-25] newpass

MPI

mpimon /opt/scali/examples/bin/hello — all

-all2all
-Crisscross
-Test4
-Pallas
-Confidence
-Stream
-Linpack

Remove & Add a License

/opt/scali/bin/lmutil remove AAEW-2YLO-MFTW-KADN-MFXA-KAAA-AAAC-LG7Z

/opt/scali/bin/lmutil add AAEW-2YLO-MFTW-KADN-MFXA-KAAA-AAAC-JWPT

SSH Key Cleanup

rm -f /var/lib/scali/scance/SshDiscovery.pickle

/etc/init.d/scance restart

And try again.

The ScaNCE SSH subsystem should also clean up the ~/.ssh/known_hosts file, in case keys have been entered in this file before the global ones had been collected.

Troubleshoot scamond

1) scamond-mapper doesn’t start because there is something wrong with the configuration files after you edited them. Please do the following :

cd /opt/scali/etc/scamond-mapper.d/

rm -rf *

rm -f /var/lib/scali/scance/ScaMondMapper.pickle

python /opt/scali/lib/python2.3/site-packages/subsystems/nce_scamond.py

that should leave you with a working scamond again.

2) the “gmond” errors in /var/log/messages are from Ganglia which has been installed on the system

3) scance fails because there is something wrong with the subnets or IP interface configuration/hosted services. Please send the output from :

scalimanage-cli listsubnets

scalimanage-cli listinterfaces ln003020

scalimanage-cli listhostedservices ln003020

Configure Static arp

If you want to revert back to dynamic ARP, all you have to do is :

scalimanage-cli disablestaticarp all eth0

scalimanage-cli disablestaticarp all eth1

scalimanage-cli reconfigure all

and likewise, if you want to turn it back on :

scalimanage-cli enablestaticarp all eth0

scalimanage-cli enablestaticarp all eth1

scalimanage-cli reconfigure all

Show BMC

scalimanage-cli showbmc [nodename]