Make sure that scoopd
is running on each compute host. If the
SCore software was installed from binary RPMs, then the script
scoopd
should be available on each compute host.
In order to start the agents, execute the following ON EACH COMPUTE HOST:
For Red Hat Linux:
# /etc/rc.d/init.d/scoopd startFor SuSE Linux:
# /sbin/init.d/scoopd start
On the server host, there should be the following rc
file:
For Red Hat Linux:
/etc/rc.d/init.d/ScoopServerFor SuSE Linux:
/etc/rc.configEdit the file and configure the following lines for your environment:
SCBDSERV=localhost SCBDGROUP=pcc SCOOPOPTIONS="-M -L"
SCBDSERV
refers to the name of the host running the cluster
database server,
scoreboard
(8), and SCBDGROUP
refers to the name
of the score group defined in the cluster database. SCOOPOPTIONS
allows for options to be specified on the ScoopServer
command
line in the script. -M
is used to disable locking the Compute
Host Lock Server ,and -L
is used to provide one-minute load
average data.
Then start the server:
For Red Hat Linux:
# /etc/rc.d/init.d/ScoopServer startFor SuSE Linux:
# /sbin/init.d/ScoopServer startIf you wish, you can execute the
ScoopServer
program directly
from the command line, and provide command-line options. Here is an example:
# /opt/score/sbin/ScoopServer -s localhost -g pcc
ScoopServer
is a wrapper script for executing the
ScoopServer
class.
ScoopServer
will collect information from the group of hosts
given by the pcc
group known by the
scoreboard
running on host localhost
.
The same command can also be achieved as follows:
# export PATH=$PATH:/opt/score/sbin # export SCBDSERV=localhost # ScoopServer -g pcc
On a Java-runnable system, execute the Scoop
Java program as
follows (make sure you have set your DISPLAY environment variable and you
can display X Window clients on your terminal):
# /opt/score/sbin/Scoop -s hostThe main Scoop GUI monitoring tool X Window will be displayed on your screen. For an example, follow this link
Scoop
is a wrapper script for executing the Scoop
class.
host
is a comma-separated list
of hosts running
ScoopServer
programs. If you will execute Scoop
on the same host as ScoopServer
, then you can use
localhost
for the name of the host.
Scoop
will display compute hosts from one or more clusters as
a block matrix. Each block represents one host in a cluster. If the
-L
option is used on the ScoopServer
command line
(see later) then the color will represent the load average during the last
minute. The
information is updated periodically. The time interval can be specified with
the -i
option on the Scoop
command line. The
default is 60 seconds.
Host resource information information can be displayed by selecting "Open Resource Information..." from the "Control" pull-down menu.
You can execute the Java commands manually. If the SCore software
was installed from binary RPMs then the JRE (Java Runtime Environment) will
already be installed under /opt/score/java/linux
. You can execute
the ScoopServer
and Scoop
commands directly on any
host installed with the path /opt/score/java/linux
and running
at least a glibc-2.1.2 based system with a recent Linux 2.2.xx kernel.
Here is an example:
# export PATH=$PATH:/opt/score/java/linux/bin # export CLASSPATH=/opt/score/lib/java:/opt/score/lib/java/scoop.jar # java ScoopServer -s localhost -g pcc -M -L & # java Scoop -s localhost
ScoopServer
will be executed in the background to collect
information from the group of hosts given by the
pcc
group known by the scoreboard
running on host localhost
. The
-M
option is used to disable locking the
Compute Host Lock Server. The -L
option is
used to collect one-minute load average data.
Scoop
will display information from the ScoopServer
just initiated, i.e. in this case, running on host
localhost
.