xcat-core/xCAT/postscripts/install_lsf

286 lines
9.0 KiB
Plaintext
Raw Permalink Normal View History

2014-11-13 06:10:55 +00:00
#!/bin/bash
#README################################################################
# (1)Check you have LSF installer script package and LSF distribution packages. e.g. "lsf9.1.3_lsfinstall.tar.Z" and "lsf9.1.3_lnx26-lib23-ppc64le.tar.Z"
# (2)Get LSF entitlement file for the edition you are installing. e.g. "platform_lsf_std_entitlement.dat"
# (3)Prepare a install.config file, install.config should be in the same directory with install_lsf and lsf_startup scripts.
# The format of install.config file
# cat install.config
# LSF_TOP=""
# LSF_ADMINS=""
# LSF_CLUSTER_NAME=""
# LSF_MASTER_LIST=""
# LSF_ENTITLEMENT_FILE="NEED A FULL PATH OF THE FILE"
# LSF_TARDIR=""
# (4)Run this script on one compute node,you can also use "updatenode <nodename> -P install_lsf" to execute this script on one compute node
# (It's not necessary to run it on each compute node).
#
# NOTE for install.config:
# -----------------
# LSF_TOP="/usr/share/lsf"
# -----------------
# Full path to the top-level installation directory {REQUIRED}
#
# The path to LSF_TOP must be shared and accessible to all hosts
# in the cluster. It cannot be the root directory (/).
# The file system containing LSF_TOP must have enough disk space for
# all host types (approximately 300 MB per host type).
# -----------------
# LSF_ADMINS="lsfadmin user1 user2"
# -----------------
# List of LSF administrators {REQUIRED}
#
# The first user account name in the list is the primary LSF
# administrator. It cannot be the root user account.
# Typically, this account is named lsfadmin.
# It owns the LSF configuration files and log files for job events.
# It also has permission to reconfigure LSF and to control batch
# jobs submitted by other users. It typically does not have
# authority to start LSF daemons. Usually, only root has
# permission to start LSF daemons.
# All the LSF administrator accounts must exist on all hosts in the
# cluster before you install LSF.
# Secondary LSF administrators are optional.
#
# -----------------
# LSF_CLUSTER_NAME="cluster1"
# -----------------
# Name of the LSF cluster {REQUIRED}
#
# It must be 39 characters or less, and cannot contain any
# white spaces. Do not use the name of any host, user, or user group
# as the name of your cluster.
#
# -----------------
# LSF_MASTER_LIST="hostm hosta hostc"
# -----------------
# List of LSF server hosts to be master or master candidate in the
# cluster {REQUIRED when you install for the first time or during
# upgrade if the parameter does not already exist.}
#
# You must specify at least one valid server host to start the
# cluster. The first host listed is the LSF master host.
#
# -----------------
# LSF_ENTITLEMENT_FILE="/usr/share/lsf/lsf_distrib/platform_lsf_std_entitlement.dat"
# -----------------
# You must specify a full path to the LSF entitlement file.
#
# -----------------
# LSF_TARDIR="/usr/share/lsf_distrib/"
# -----------------
# Full path to the directory containing the LSF distribution tar files.
#
# Default: Parent directory of the current working directory.
# For example, if lsfinstall is running under
# /usr/share/lsf_distrib/lsf_lsfinstall
# the LSF_TARDIR default value is
# /usr/share/lsf_distrib
# -----------------
# LSF_ADD_SERVERS="hostm hosta hostb hostc"
# -----------------
# List of additional LSF server hosts
#
# The hosts in LSF_MASTER_LIST are always LSF servers. You can specify
# additional server hosts.
#README################################################################
INSTALL_CONFIG_FILE=`pwd`/install.config
#LSF_INSTALL_FILE_PATH=""
#LSF_GLIBC_FILE_PATH=""
function is_parameter_set()
{
PARA=$1
IF_SET=`grep $PARA $INSTALL_CONFIG_FILE`
if [[ -z $IF_SET ]] ; then
if [[ $PARA == "LSF_SILENT_INSTALL_TARLIST" ]]; then
echo "$PARA="all"" >> $INSTALL_CONFIG_FILE
elif [[ $PARA == "LSF_DYNAMIC_HOST_WAIT_TIME" ]]; then
echo "$PARA="60"" >> $INSTALL_CONFIG_FILE
else
echo "$PARA="Y"" >> $INSTALL_CONFIG_FILE
fi
echo "INFO: Set a recommended value for $PARA"
fi
return 0
}
#verify if install.config exists
if [[ ! -f ${INSTALL_CONFIG_FILE} ]]; then
echo "ERROR: $INSTALL_CONFIG_FILE not found"
exit 1
fi
. $INSTALL_CONFIG_FILE
cat $INSTALL_CONFIG_FILE
#verify if the required parameters are valid in install.config
if [[ ! -d $LSF_TOP ]]; then
echo "ERROR: No such directory $LSF_TOP. Check LSF_TOP in install.config"
exit 1
fi
if [[ x$LSF_ADMINS == x ]]; then
echo "ERROR: You must specify LSF_ADMINS in install.config"
exit 1
fi
for USERID in ${LSF_ADMINS} ; do
RTC=`id $USERID`
RTC=$?
if [[ $RTC -ne 0 ]] ; then
echo "ERROR: No such user id $USERID on the node. Check LSF_ADMINS in install.config"
exit 1
fi
done
if [[ x$LSF_CLUSTER_NAME == x ]]; then
echo "ERROR: You must specify LSF_CLUSTER_NAME in install.config"
exit 1
fi
if [[ x$LSF_MASTER_LIST == x ]]; then
echo "ERROR: You must specify LSF_MASTER_LIST in install.config"
exit 1
fi
for MASTER_NODE in $LSF_MASTER_LIST ; do
RTC=`ssh $MASTER_NODE uptime`
RTC=$?
if [[ $RTC -ne 0 ]] ; then
echo "ERROR: MASTER_NODE $MASTER_NODE is not reachable. Check LSF_MASTER_LIST in install.config"
exit 1
fi
done
if [[ ! -r ${LSF_ENTITLEMENT_FILE} || -d ${LSF_ENTITLEMENT_FILE} ]]; then
echo "ERROR: No such entitlement file $LSF_ENTITLEMENT_FILE. Check LSF_ENTITLEMENT_FILE in install.config"
exit 1
fi
#CH_TMP=`expr "$LSF_ENTITLEMENT_FILE" : '\(.\).*'`
#if [[ "$CH_TMP" != "/" ]]; then
# echo "Set LSF_ENTITLEMENT_FILE a full path but not relative path"
# return 1
#fi
if [[ ! -d $LSF_TARDIR ]]; then
echo "ERROR: No such directory $LSF_TARDIR. Check LSF_TARDIR in install.config"
exit 1
fi
if [[ x$LSF_INSTALL_FILE_PATH != x ]]; then
if [[ -r ${LSF_INSTALL_FILE_PATH} ]] && [[ ! -d ${LSF_INSTALL_FILE_PATH} ]]; then
:
else
echo "ERROR: No such lsf install tar file $LSF_INSTALL_FILE_PATH. Check LSF_INSTALL_FILE_PATH in install.config"
exit 1
fi
else
LSF_INSTALL_FILE_PATH=`find $LSF_TARDIR -name 'lsf*lsfinstall*tar.Z'`
count=`echo $LSF_INSTALL_FILE_PATH |wc -w`
if [[ $count -gt 1 ]]; then
echo "ERROR: There are more than one lsfinstall tar file. You need to specify LSF_INSTALL_FILE_PATH in install.config or remove other useless lsfinstall TAR files in LSF_TARDIR $LSF_TARDIR."
exit 1
elif [[ $count -eq 1 ]]; then
:
else
echo "ERROR: lsfinstall TAR file not found in LSF_TARDIR $LSF_TARDIR."
exit 1
fi
fi
echo "INFO: We will untar the lsfinstall TAR file $LSF_INSTALL_FILE_PATH."
for SERVER_NODE in $LSF_ADD_SERVERS ; do
RTC=`ssh $SERVER_NODE uptime`
RTC=$?
if [[ $RTC -ne 0 ]] ; then
echo "ERROR: SERVER_NODE $SERVER_NODE is not reachable. Check LSF_ADD_SERVERS in install.config"
exit 1
fi
done
#Check if we set following parameters in install.config; if not, set them a recommended value.
#ENABLE_DYNAMIC_HOSTS="Y"
##LSF_DYNAMIC_HOST_WAIT_TIME="60"
#ENABLE_HPC_CONFIG="Y"
#SILENT_INSTALL="Y"
#LSF_SILENT_INSTALL_TARLIST="all"
is_parameter_set "ENABLE_DYNAMIC_HOSTS"
is_parameter_set "LSF_DYNAMIC_HOST_WAIT_TIME"
is_parameter_set "ENABLE_HPC_CONFIG"
is_parameter_set "SILENT_INSTALL"
is_parameter_set "LSF_SILENT_INSTALL_TARLIST"
#Extract lsfinstall package
cd $LSF_TARDIR
LSF_INSTALL_PACKAGE=`basename $LSF_INSTALL_FILE_PATH`
zcat $LSF_INSTALL_PACKAGE | tar xvf -
RTC=$?
if [[ $RTC -ne 0 ]] ; then
echo "ERROR: Fail to extract LSF_INSTALL_PACKAGE $LSF_INSTALL_PACKAGE."
exit 1
fi
#INSTALL LSF on the node
cd $LSF_TARDIR/lsf*lsfinstall
INSTALL_LOG="Install.log"
#backup Install.log if there is one before installing
if [[ -r $INSTALL_LOG ]]; then
mv $INSTALL_LOG `date "+%Y.%m.%d-%H:%M"`_$INSTALL_LOG
fi
./lsfinstall -f $INSTALL_CONFIG_FILE
RTC=$?
IF_INSTALL_DONE=`grep "lsfinstall is done" $INSTALL_LOG`
if [[ $RTC -ne 0 || -z $IF_INSTALL_DONE ]] ; then
echo "ERROR: Fail to install LSF. Check Install.log and Install.err in `pwd`."
exit 1
fi
echo "INFO: Installation script DONE."
LSF_VERSION=`echo $LSF_INSTALL_PACKAGE |cut -c4-6`
if [[ `echo "$LSF_VERSION >= 9.1"|bc` -eq 1 ]]
then
#Start configuration. Update configuration files lsf.conf,lsf.hosts.
echo "INFO: Updating LSF Cluster Configuration Files lsf.conf and lsb.hosts"
LSF_CONF_FILE="$LSF_TOP/conf/lsf.conf"
echo "LSF_RSH=ssh" >> $LSF_CONF_FILE
echo "LSF_PE_NETWORK_NUM=2" >> $LSF_CONF_FILE
echo "LSF_PE_NETWORK_UPDATE_INTERVAL=6" >> $LSF_CONF_FILE
echo "EGO_DEFINE_NCPUS=threads" >> $LSF_CONF_FILE
echo "LSF_HPC_EXTENSIONS="CUMULATIVE_RUSAGE"">> $LSF_CONF_FILE
LSB_HOSTS_FILE="$LSF_TOP/conf/lsbatch/$LSF_CLUSTER_NAME/configdir/lsb.hosts"
LSB_HOSTS_FILE_ORIG="$LSF_TOP/conf/lsbatch/$LSF_CLUSTER_NAME/configdir/lsb.hosts.orig"
mv $LSB_HOSTS_FILE $LSB_HOSTS_FILE_ORIG
PRIMARY_MASTER_NODE=`echo $LSF_MASTER_LIST | awk '{print $1}'`
echo "Begin Host" >> lsb.hosts
echo "HOST_NAME MXJ r1m pg ls tmp DISPATCH_WINDOW AFFINITY" >> $LSB_HOSTS_FILE
echo "default ! () () () () () (Y)" >> $LSB_HOSTS_FILE
echo "$PRIMARY_MASTER_NODE 0 () () () () () (Y)" >> $LSB_HOSTS_FILE
echo "End Host" >> $LSB_HOSTS_FILE
echo "Begin HostGroup" >> $LSB_HOSTS_FILE
echo "GROUP_NAME GROUP_MEMBER" >> $LSB_HOSTS_FILE
echo "End HostGroup" >> $LSB_HOSTS_FILE
fi