lustre_createcsv.in

来自「lustre 1.6.5 source code」· IN 代码 · 共 2,102 行 · 第 1/4 页

#!/bin/bash
#
# lustre_createcsv - generate a csv file from a running lustre cluster
#
# This script is used to collect lustre target informations, linux MD/LVM device
# informations and HA software configurations in a lustre cluster to generate a
# csv file. In reverse, the csv file could be parsed by lustre_config to 
# configure multiple lustre servers in parallel.
#
# This script should be run on the MGS node.
#
################################################################################

# Usage
usage() {
	cat >&2 <<EOF

Usage:	`basename $0` [-t HAtype] [-d] [-h] [-v] [-f csv_filename]

	This script is used to collect lustre target informations, linux MD/LVM
	device informations and HA software configurations from a running lustre
	cluster to generate a csv file. It should be run on the MGS node.

	-t HAtype	collect High-Availability software configurations
			The argument following -t is used to indicate the High-
			Availability software type. The HA software types which 
			are currently supported are: hbv1 (Heartbeat version 1)
			and hbv2 (Heartbeat version 2).
	-d		collect linux MD/LVM device informations
	-h		help
	-v		verbose mode
	-f csv_filename	designate a name for the csv file
			Default is lustre_config.csv.

EOF
	exit 1
}

# Get the library of functions
. @scriptlibdir@/lc_common

#**************************** Global variables ****************************#
# csv file
LUSTRE_CSV_FILE=${LUSTRE_CSV_FILE:-"lustre_config.csv"}

# Lustre proc files
LUSTRE_PROC=${LUSTRE_PROC:-"/proc/fs/lustre"}
LUSTRE_PROC_DEVICES=${LUSTRE_PROC}/devices

LNET_PROC=${LNET_PROC:-"/proc/sys/lnet"}
LNET_PROC_PEERS=${LNET_PROC}/peers

# Default network module options
DEFAULT_MOD_OPTS=${DEFAULT_MOD_OPTS:-"options lnet networks=tcp"}

# Lustre target obd device types
MGS_TYPE=${MGS_TYPE:-"mgs"}
MDT_TYPE=${MDT_TYPE:-"mds"}
OST_TYPE=${OST_TYPE:-"obdfilter"}

# The obd name of MGS target server
MGS_SVNAME=${MGS_SVNAME:-"MGS"}		

# Hostnames of the lustre cluster nodes
declare -a HOST_NAMES			
MGS_HOSTNAME=${MGS_HOSTNAME:-"`hostname`"} # Hostname of the MGS node

# Configs of lustre targets in one cluster node
declare -a TARGET_CONFIGS		
declare -a TARGET_SVNAMES TARGET_DEVNAMES TARGET_DEVSIZES TARGET_MNTPNTS
declare -a TARGET_DEVTYPES TARGET_FSNAMES TARGET_MGSNIDS TARGET_INDEXES
declare -a TARGET_FMTOPTS TARGET_MKFSOPTS TARGET_MNTOPTS TARGET_FAILNIDS
declare -a HA_CONFIGS
declare -a ALL_TARGET_SVNAMES		# All the target services in the cluster
declare -a FAILOVER_FMTOPTS		# "--noformat"	

# Informations of linux MD/LVM devices in one cluster node
declare -a MD_NAME MD_LEVEL MD_DEVS	# MD
declare -a VG_NAME VG_PVNAMES		# VG
declare -a LV_NAME LV_SIZE LV_VGNAME	# LV

# Lustre target service types
let "LDD_F_SV_TYPE_MDT = 0x0001"
let "LDD_F_SV_TYPE_OST = 0x0002"
let "LDD_F_SV_TYPE_MGS = 0x0004"

# Permanent mount options for ext3 or ldiskfs
ALWAYS_MNTOPTS=${ALWAYS_MNTOPTS:-"errors=remount-ro"}
MDT_MGS_ALWAYS_MNTOPTS=${MDT_MGS_ALWAYS_MNTOPTS:-",iopen_nopriv,user_xattr"}
OST_ALWAYS_MNTOPTS=${OST_ALWAYS_MNTOPTS:-",asyncdel"}
OST_DEFAULT_MNTOPTS=${OST_DEFAULT_MNTOPTS:-",extents,mballoc"}

# User-settable parameter keys
PARAM_MGSNODE=${PARAM_MGSNODE:-"mgsnode="}
PARAM_FAILNODE=${PARAM_FAILNODE:-"failover.node="}

# Block size
L_BLOCK_SIZE=4096

# Option string of mkfs.lustre
OPTSTR_STRIPE_COUNT=${OPTSTR_STRIPE_COUNT:-"--stripe-count-hint="}


# Get and check the positional parameters
VERBOSE_OUTPUT=false
GET_MDLVM_INFO=false
while getopts "t:dhvf:" OPTION; do
	case $OPTION in
	t) 
		HATYPE_OPT=$OPTARG
		if [ "${HATYPE_OPT}" != "${HBVER_HBV1}" ] \
		&& [ "${HATYPE_OPT}" != "${HBVER_HBV2}" ] \
		&& [ "${HATYPE_OPT}" != "${HATYPE_CLUMGR}" ]; then
			echo >&2 "`basename $0`: Invalid HA software type" \
				 "- ${HATYPE_OPT}!"
			usage
		fi
		;;
	d)	GET_MDLVM_INFO=true;;
	h)	usage;;
	v) 	VERBOSE_OUTPUT=true;;
	f)	LUSTRE_CSV_FILE=$OPTARG;;
        ?) 	usage 
	esac
done

# Verify the local host is the MGS node
mgs_node() {
	if [ ! -e ${LUSTRE_PROC_DEVICES} ]; then
		echo >&2 "`basename $0`: error: ${LUSTRE_PROC_DEVICES} does" \
			 "not exist. Lustre kernel modules may not be loaded!"
		return 1
	fi

	if [ -z "`cat ${LUSTRE_PROC_DEVICES}`" ]; then
		echo >&2 "`basename $0`: error: ${LUSTRE_PROC_DEVICES} is" \
			 "empty. Lustre services may not be started!"
		return 1
	fi

	if [ -z "`grep ${MGS_TYPE} ${LUSTRE_PROC_DEVICES}`" ]; then
		echo >&2 "`basename $0`: error: This node is not a MGS node." \
                         "The script should be run on the MGS node!"
		return 1
	fi

	return 0
}

# get_hostnames
# Get lustre cluster node names
get_hostnames() {
	declare -a HOST_NIDS
	declare -i idx		# Index of HOST_NIDS array
	declare -i i		# Index of HOST_NAMES array

	if ! mgs_node; then
		return 1
	fi

	if [ ! -e ${LNET_PROC_PEERS} ]; then
		echo >&2 "`basename $0`: error: ${LNET_PROC_PEERS} does not" \
                         "exist. LNET kernel modules may not be loaded" \
			 "or LNET network may not be up!"
		return 1
	fi

	HOST_NAMES[0]=${MGS_HOSTNAME} # MGS node
	HOST_NIDS[0]=${HOST_NAMES[0]}

	# Get the nids of the nodes which have contacted MGS
	idx=1
	for nid in `cat ${LNET_PROC_PEERS} | awk '{print $1}'`; do
		if [ "${nid}" = "nid" ]; then
			continue
		fi

		HOST_NIDS[idx]=${nid}
		let "idx += 1"
	done

	if [ ${idx} -eq 1 ]; then
		verbose_output "Only one node running in the lustre cluster." \
			       "It's ${HOST_NAMES[0]}."
		return 0		
	fi

	# Get the hostnames of the nodes
	for ((idx = 1, i = 1; idx < ${#HOST_NIDS[@]}; idx++, i++)); do
		if [ -z "${HOST_NIDS[idx]}" ]; then
			echo >&2 "`basename $0`: get_hostnames() error:" \
				 "Invalid nid - \"${HOST_NIDS[idx]}\"!"
			return 1
		fi

		HOST_NAMES[i]=$(nid2hostname ${HOST_NIDS[idx]})
		if [ $? -ne 0 ]; then
			echo >&2 "${HOST_NAMES[i]}"
			return 1
		fi

		if [ "${HOST_NAMES[i]}" = "${HOST_NAMES[0]}" ]; then
			unset HOST_NAMES[i]
			let "i -= 1"
		fi
	done

	return 0
}

#********************** Linux MD/LVM device informations **********************#
# get_md_configs hostname
# Get all the active MD device informations from the node @hostname
get_md_configs() {
	declare -i i=0
	declare -i j=0
	local host_name=$1
	local ret_line line first_item

	# Initialize the arrays
	unset MD_NAME
	unset MD_LEVEL
	unset MD_DEVS
	
	# Execute remote command to the node ${host_name} and get all the
	# active MD device informations.
	while read -r ret_line; do
		if is_pdsh; then
                	set -- ${ret_line}
			shift
			line="$*"
		else
			line="${ret_line}"
		fi

		first_item=`echo "${line}" | awk '{print $1}'`

		# Get the MD device name and raid level
		if [ "${first_item}" = "ARRAY" ]; then
			MD_NAME[i]=`echo "${line}" | awk '{print $2}'`
			MD_LEVEL[i]=`echo "${line}" | awk '{print $3}' | sed -e 's/level=//'`
			let "j = i"
			let "i += 1"
		fi

		# Get the MD component devices
		if [ "${first_item}" != "${first_item#devices=}" ]; then
			MD_DEVS[j]=`echo "${line}" | sed -e 's/devices=//' -e 's/,/ /g'`
		fi
        done < <(${REMOTE} ${host_name} "${MDADM} --detail --scan --verbose")

	if [ $i -eq 0 ]; then
		verbose_output "There are no active MD devices" \
			       "in the host ${host_name}!"
	fi

	return 0
}

# get_pv_configs hostname
# Get all the LVM PV informations from the node @hostname
get_pv_configs() {
	PV_NAMES=
	local host_name=$1
	local cmd ret_str

	# Execute remote command to get all the PV informations.
	cmd="${EXPORT_PATH} pvdisplay -c | awk -F: '{print \$1}' | xargs"
	ret_str=`${REMOTE} ${host_name} "${cmd}" 2>&1`
	if [ $? -ne 0 ]; then
		if [ -n "${ret_str}" ]; then
			echo >&2 "`basename $0`: get_pv_configs() error:" \
			"remote command to ${host_name} error: ${ret_str}"
		else
			remote_error "get_pv_configs" ${host_name}
		fi
		return 1
	fi

	PV_NAMES=`echo "${ret_str}" | sed -e 's/^'${host_name}':[[:space:]]//'`
	if [ -z "${PV_NAMES}" ]; then
		verbose_output "There are no PVs in the host ${host_name}!"
		return 0
	fi

	return 0
}

# get_vg_pvnames hostname vgname
# Get the PVs contained in @vgname from the node @hostname
get_vg_pvnames() {
	local host_name=$1
	local vg_name=$2
	local pv_names=
	local cmd ret_str

	# Execute remote command to get the PV names.
	cmd="${EXPORT_PATH} vgdisplay -v ${vg_name} 2>/dev/null\
	     | grep \"PV Name\" | awk '{print \$3}' | xargs"
	ret_str=`${REMOTE} ${host_name} "${cmd}" 2>&1`
	if [ $? -ne 0 ]; then
		if [ -n "${ret_str}" ]; then
			echo "`basename $0`: get_vg_pvnames() error:" \
			"remote command to ${host_name} error: ${ret_str}"
		else
			remote_error "get_vg_pvnames" ${host_name}
		fi
		return 1
	fi

	pv_names=`echo "${ret_str}" | sed -e 's/^'${host_name}':[[:space:]]//'`
	if [ -z "${pv_names}" ]; then
		echo "`basename $0`: get_vg_pvnames() error:" \
		"There are no PVs in VG ${vg_name} in the host ${host_name}!"\
		"Or VG ${vg_name} does not exist."
		return 1
	fi

	echo "${pv_names}"
	return 0
}

# get_vg_configs hostname
# Get all the LVM VG informations from the node @hostname
get_vg_configs() {
	declare -i i=0
	local host_name=$1
	local cmd ret_str
	local vg_name

	# Initialize the arrays
	unset VG_NAME
	unset VG_PVNAMES

	# Execute remote command to get all the VG names.
	cmd="${EXPORT_PATH} vgdisplay \
	     | grep \"VG Name\" | awk '{print \$3}' | xargs"
	ret_str=`${REMOTE} ${host_name} "${cmd}" 2>&1`
	if [ $? -ne 0 ]; then
		if [ -n "${ret_str}" ]; then
			echo >&2 "`basename $0`: get_vg_configs() error:" \
			"remote command to ${host_name} error: ${ret_str}"
		else
			remote_error "get_vg_configs" ${host_name}
		fi
		return 1
	fi

	if [ -z "${ret_str}" ] \
	|| [ "${ret_str}" != "${ret_str#*No volume groups found*}" ]; then
		verbose_output "There are no VGs in the host ${host_name}!"
		return 0
	fi

	# Get all the VG informations
	for vg_name in `echo "${ret_str}" | sed -e 's/^'${host_name}'://'`; do
		VG_NAME[i]=${vg_name}
		VG_PVNAMES[i]=$(get_vg_pvnames ${host_name} ${VG_NAME[i]})
		if [ $? -ne 0 ]; then
			echo >&2 "${VG_PVNAMES[i]}"
			return 1
		fi
		let "i += 1"
	done

	return 0
}

# get_lv_configs hostname
# Get all the LVM LV informations from the node @hostname
get_lv_configs() {
	declare -i i=0
	local host_name=$1
	local ret_line line

	# Initialize the arrays
	unset LV_NAME
	unset LV_SIZE
	unset LV_VGNAME

	# Execute remote command to get all the LV informations.
	while read -r ret_line; do
		if is_pdsh; then
                	set -- ${ret_line}
			shift
			line="$*"
		else
			line="${ret_line}"
		fi

		[ "${line}" != "${line#*volume group*}" ] && break

		LV_NAME[i]=`echo "${line}" | awk -F: '{print $1}' | sed -e 's/.*\///g'`
		LV_VGNAME[i]=`echo "${line}" | awk -F: '{print $2}'`
		LV_SIZE[i]=`echo "${line}" | awk -F: '{print $7}' | sed -e 's/.*/&K/'`

		let "i += 1"
        done < <(${REMOTE} ${host_name} "${EXPORT_PATH} lvdisplay -c")

	if [ $i -eq 0 ]; then
		verbose_output "There are no LVs in the host ${host_name}"
	fi

	return 0
}

#*************************** Network module options ***************************#
# last_is_backslash line
# Check whether the last effective letter of @line is a backslash
last_is_backslash() {
	local line="$*"
	declare -i i
	declare -i length
	local letter last_letter

	length=${#line}
	for ((i = ${length}-1; i >= 0; i--)); do
		letter=${line:${i}:1}
        	[ "x${letter}" != "x " -a "x${letter}" != "x	" -a -n "${letter}" ]\
		&& last_letter=${letter} && break
	done

	[ "x${last_letter}" = "x\\" ] && return 0

	return 1
}

# get_module_opts hostname
# Get the network module options from the node @hostname 
get_module_opts() {
	local host_name=$1
	local ret_str
	local MODULE_CONF KERNEL_VER
	local ret_line line find_options
	local continue_flag

	MODULE_OPTS=${DEFAULT_MOD_OPTS}

	# Execute remote command to get the kernel version
	ret_str=`${REMOTE} ${host_name} "uname -r" 2>&1`
	if [ $? -ne 0 -a -n "${ret_str}" ]; then
		echo >&2 "`basename $0`: get_module_opts() error:" \
			 "remote command error: ${ret_str}"
		return 1
	fi
	remote_error "get_module_opts" ${host_name} "${ret_str}" && return 1

	if is_pdsh; then
		KERNEL_VER=`echo ${ret_str} | awk '{print $2}'`
	else
		KERNEL_VER=`echo ${ret_str} | awk '{print $1}'`
	fi

	# Get the module configuration file name
	if [ "${KERNEL_VER:0:3}" = "2.4" ]; then
        	MODULE_CONF=/etc/modules.conf
	else
        	MODULE_CONF=/etc/modprobe.conf
	fi

	# Execute remote command to get the lustre network module options
	continue_flag=false
	find_options=false
	while read -r ret_line; do
		if is_pdsh; then
                	set -- ${ret_line}
			shift
			line="$*"
		else
			line="${ret_line}"
		fi

		# Get rid of the comment line
		[ -z "`echo \"${line}\"|egrep -v \"^#\"`" ] && continue

		if [ "${line}" != "${line#*options lnet*}" ]; then
			if ! ${find_options}; then
				find_options=true
				MODULE_OPTS=${line}
			else
				MODULE_OPTS=${MODULE_OPTS}$" \n "${line}
			fi

			last_is_backslash "${line}" && continue_flag=true \
			|| continue_flag=false
			continue
		fi	

		if ${continue_flag}; then
			MODULE_OPTS=${MODULE_OPTS}$" \n "${line}
			! last_is_backslash "${line}" && continue_flag=false

		fi
        done < <(${REMOTE} ${host_name} "cat ${MODULE_CONF}")

	if [ -z "${MODULE_OPTS}" ]; then
		MODULE_OPTS=${DEFAULT_MOD_OPTS}
	fi

	return 0
}

#************************ HA software configurations ************************#
# is_ha_target hostname target_devname
# Check whether the target @target_devname was made to be high-available
is_ha_target() {
	local host_name=$1
	local target_svname=$2
	local res_file
	local ret_str

	case "${HATYPE_OPT}" in
	"${HBVER_HBV1}")	res_file=${HA_RES};;
	"${HBVER_HBV2}")	res_file=${HA_CIB};;
	"${HATYPE_CLUMGR}")	res_file=${CLUMAN_CONFIG};;
	esac

	# Execute remote command to check the resource file
	ret_str=`${REMOTE} ${host_name} \
		"grep ${target_svname} ${res_file}" 2>&1`
	if [ $? -ne 0 -a -n "${ret_str}" ]; then
		echo >&2 "`basename $0`: is_ha_target() error:" \
			 "remote command error: ${ret_str}"
		return 1
	fi