gp_bash_functions.sh 42.6 KB
Newer Older
1
#!/usr/bin/env bash
2
#	Filename:-		gp_bash_functions.sh
D
Daniel Gustafsson 已提交
3
#	Status:-		Released
4 5 6 7 8 9 10 11 12 13
#	Author:-		G L Coombe (Greenplum)
#	Contact:-		gcoombe@greenplum.com
#	Release date:-		March 2006
#	Release stat:-		Greenplum Internal
#                               Copyright (c) Metapa 2005. All Rights Reserved.
#                               Copyright (c) Greenplum 2005. All Rights Reserved
#	Brief descn:-		Common functions used by various scripts
#***************************************************************
# Location Functions
#******************************************************************************
14 15 16
#Check that SHELL is Bash
if [ -z $BASH ]; then
	echo "[FATAL]:-Scripts must be executed using the Bash shell"
S
Shoaib Lari 已提交
17
	exit 1
18
fi
19 20 21 22 23 24 25 26
#CMDPATH is the list of locations to search for commands, in precedence order
declare -a CMDPATH
CMDPATH=(/usr/kerberos/bin /usr/sfw/bin /opt/sfw/bin /usr/local/bin /bin /usr/bin /sbin /usr/sbin /usr/ucb /sw/bin)

#GPPATH is the list of possible locations for the Greenplum Database binaries, in precedence order
declare -a GPPATH
GPPATH=( $GPHOME $MPPHOME $BIZHOME )
if [ ${#GPPATH[@]} -eq 0 ];then
27 28 29
	echo "[FATAL]:-GPHOME environment variable is required to run GPDB but could not be found."
	echo "Please set it by sourcing the  greenplum_path.sh  in your GPDB installation directory."
	echo "Example: ''. /usr/local/gpdb/greenplum_path.sh''"
S
Shoaib Lari 已提交
30
	exit 1
31 32 33 34 35 36 37 38
fi

#GP_UNIQUE_COMMAND is used to identify the binary directory
GP_UNIQUE_COMMAND=gpstart


findCmdInPath() {
		cmdtofind=$1
D
Daniel Gustafsson 已提交
39

40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114
		for pathel in ${CMDPATH[@]}
				do
				CMD=$pathel/$cmdtofind
				if [ x"$CMD" != x"" ] && [ -f $CMD ]; then
						echo $CMD
						return
				fi
		done
		echo $cmdtofind
		return "Problem in gp_bash_functions, command '$cmdtofind' not found in COMMAND path. You will need to edit the script named gp_bash_functions.sh to properly locate the needed commands for your platform."
}

findMppPath() {
		cmdtofind=$GP_UNIQUE_COMMAND

		for pathel in ${GPPATH[@]}
				do
				CMD=`find $pathel -follow -name $cmdtofind | tail -1`
				if [ x"$CMD" != x"" ] && [ -f $CMD ]; then
						echo $CMD
						return
				fi
		done
}

#******************************************************************************
# OS Command Variables
#******************************************************************************
AWK=`findCmdInPath awk`
BASENAME=`findCmdInPath basename`
CAT=`findCmdInPath cat`
CUT=`findCmdInPath cut`
DATE=`findCmdInPath date`
DIRNAME=`findCmdInPath dirname`
ECHO=`findCmdInPath echo`
FIND=`findCmdInPath find`
GREP=`findCmdInPath grep`
EGREP=`findCmdInPath egrep`
HEAD=`findCmdInPath head`
HOSTNAME=`findCmdInPath hostname`
IFCONFIG=`findCmdInPath ifconfig`
LESSCMD=`findCmdInPath less`
LOCALE=`findCmdInPath locale`
MV=`findCmdInPath mv`
MKDIR=`findCmdInPath mkdir`
NETSTAT=`findCmdInPath netstat`
PING=`findCmdInPath ping`
RM=`findCmdInPath rm`
SCP=`findCmdInPath scp`
SED=`findCmdInPath sed`
SLEEP=`findCmdInPath sleep`
SORT=`findCmdInPath sort`
SSH=`findCmdInPath ssh`
TAIL=`findCmdInPath tail`
TEE=`findCmdInPath tee`
TOUCH=`findCmdInPath touch`
TR=`findCmdInPath tr`
WC=`findCmdInPath wc`
#***************#******************************************************************************
# Script Specific Variables
#******************************************************************************
# By default set error logging level to verbose
VERBOSE=1
USER_NAME=`id|$AWK '{print $1}'|$CUT -d"(" -f2|$TR -d ')'`
PROG_NAME=`echo $0 | $TR -d '-'`
PROG_NAME=`$BASENAME $PROG_NAME`
PROG_PIDNAME=`echo $$ $PROG_NAME | awk '{printf "%06d %s\n", $1, $2}'`
CALL_HOST=`$HOSTNAME|$CUT -d. -f1`

#******************************************************************************
# Locate the postgres routines from the Greenplum release
#******************************************************************************
PSQLBIN=`findMppPath`

if [ x"$PSQLBIN" = x"" ];then
115 116
		echo "Problem in gp_bash_functions, command '$GP_UNIQUE_COMMAND' not found in Greenplum path."
		echo "Try setting GPHOME to the location of your Greenplum distribution."
S
Shoaib Lari 已提交
117
		exit 1
118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
fi

PSQLBIN=`$DIRNAME $PSQLBIN`
SCRIPTDIR="`$DIRNAME $PSQLBIN`/bin"
#******************************************************************************
# Greenplum Scripts
#******************************************************************************
GPINITSYSTEM=$SCRIPTDIR/gpinitsystem
GPCONFIG=$SCRIPTDIR/gpconfig
GPINITSTANDBY=$SCRIPTDIR/gpinitstandby
GPRECOVERSEG=$SCRIPTDIR/gprecoverseg
GPSTART=$SCRIPTDIR/gpstart
GPSTATE=$SCRIPTDIR/gpstate
GPSTOP=$SCRIPTDIR/gpstop
GPDOCDIR=${GPHOME}/docs/cli_help/
#******************************************************************************
# Greenplum Command Variables
#******************************************************************************
INITDB=$PSQLBIN/initdb
PG_CTL=$PSQLBIN/pg_ctl
PSQL=$PSQLBIN/psql

#******************************************************************************
D
Daniel Gustafsson 已提交
141
# Greenplum OS Settings
142 143 144 145 146 147 148 149 150
#******************************************************************************
OS_OPENFILES=65535
#******************************************************************************
# General Variables
#******************************************************************************
HOSTFILE=/etc/hosts
PG_PID=postmaster.pid
PG_OPT=postmaster.opts
PG_CONF=postgresql.conf
151
PG_INTERNAL_CONF=internal.auto.conf
152 153 154 155 156 157 158 159 160 161 162 163 164 165 166 167 168 169 170 171 172 173 174 175 176 177 178 179 180 181 182
PG_HBA=pg_hba.conf
if [ x"$TRUSTED_SHELL" = x"" ]; then TRUSTED_SHELL="$SSH"; fi
if [ x"$TRUSTED_COPY" = x"" ]; then TRUSTED_COPY="$SCP"; fi
PG_CONF_ADD_FILE=$WORKDIR/postgresql_conf_gp_additions
DEFAULTDB=template1

DEFAULT_CHK_PT_SEG=8
DEFAULT_QD_MAX_CONNECT=250
QE_CONNECT_FACTOR=3
# DEFAULT_BUFFERS sets the default shared_buffers unless overridden by '-b'.
# It applies to the master db and segment dbs.  Specify either the number of
# buffers (without suffix) or the amount of memory to use for buffers (with
# case-insensitive suffix 'kB', 'MB' or 'GB').
DEFAULT_BUFFERS=128000kB
DEBUG_LEVEL=0
BATCH_DEFAULT=60
WAIT_LIMIT=1800
WARN_MARK="<<<<<"
#******************************************************************************
# Functions
#******************************************************************************

IN_ARRAY () {
    for v in $2; do
        if [ x"$1" == x"$v" ]; then
            return 1
        fi
    done
    return 0
}

183 184 185
#
# NOTE: this function is called a lot; try to keep it quick.
#
186
LOG_MSG () {
187
		TIMESTAMP=`$DATE +%Y%m%d":"%H":"%M":"%S`
188
		DISPLAY_TXT=0
189 190 191 192 193 194 195

		# Check to see if we need to update value of EXIT_STATUS. Strip off
		# everything in the message after the first ending bracket ']' and
		# compare it to WARN/FATAL.
		level=${1%%]*}
		case "$level" in
		*FATAL*)
S
Shoaib Lari 已提交
196
			EXIT_STATUS=1
197 198 199
			;;
		esac

200 201 202
		if [ x"" == x"$DEBUG_LEVEL" ];then
			DEBUG_LEVEL=1
		fi
D
Daniel Gustafsson 已提交
203
		if [ $# -eq 2 ];then
204 205 206 207
			DISPLAY_TXT=1
		fi
		if [ $VERBOSE ]; then
				if [ $DEBUG_LEVEL -eq 1 ] || [ $DISPLAY_TXT -eq 1 ];then
208
					$ECHO "${TIMESTAMP}:${PROG_PIDNAME}:${CALL_HOST}:${USER_NAME}-$1" | $TEE -a $LOG_FILE
209
				else
210
					$ECHO "${TIMESTAMP}:${PROG_PIDNAME}:${CALL_HOST}:${USER_NAME}-$1" >> $LOG_FILE
211 212
				fi
		else
213
				$ECHO "${TIMESTAMP}:${PROG_PIDNAME}:${CALL_HOST}:${USER_NAME}-$1" >> $LOG_FILE
214 215 216 217 218 219 220 221 222 223 224 225 226 227 228 229 230 231 232 233 234 235 236 237
		fi
}

POSTGRES_VERSION_CHK() {
    LOG_MSG "[INFO]:-Start Function $FUNCNAME"
    HOST=$1;shift

    CURRENT_VERSION=`$EXPORT_GPHOME; $EXPORT_LIB_PATH; $GPHOME/bin/postgres --gp-version`
    VERSION_MATCH=0

    VER=`$TRUSTED_SHELL $HOST "$EXPORT_GPHOME; $EXPORT_LIB_PATH; $GPHOME/bin/postgres --gp-version"`
    if [ $? -ne 0 ] ; then
	LOG_MSG "[WARN]:- Failed to obtain postgres version on $HOST" 1
	VERSION_MATCH=0
    fi
    LOG_MSG "[INFO]:- Current postgres version = $CURRENT_VERSION"
    LOG_MSG "[INFO]:- postgres version on $HOST = $VER"

    if [ x"$VER" != x"$CURRENT_VERSION" ] ; then
	LOG_MSG "[WARN]:-Postgres version does not match. [$CURRENT_VERSION != $VER]" 1
	VERSION_MATCH=0
    else
	VERSION_MATCH=1
    fi
D
Daniel Gustafsson 已提交
238

239 240 241 242 243 244 245 246 247

    LOG_MSG "[INFO]:-End Function $FUNCNAME"

}

ERROR_EXIT () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
		TIME=`$DATE +%H":"%M":"%S`
		CUR_DATE=`$DATE +%Y%m%d`
248 249
		$ECHO "${CUR_DATE}:${TIME}:${PROG_PIDNAME}:${CALL_HOST}:${USER_NAME}-$1 Script Exiting!" >> $LOG_FILE
		$ECHO "${CUR_DATE}:${TIME}:${PROG_PIDNAME}:${CALL_HOST}:${USER_NAME}-$1 Script Exiting!"
250 251 252 253
		DEBUG_LEVEL=1
		if [ $BACKOUT_FILE ]; then
				if [ -s $BACKOUT_FILE ]; then
						LOG_MSG "[WARN]:-Script has left Greenplum Database in an incomplete state"
254
						LOG_MSG "[WARN]:-Run command bash $BACKOUT_FILE to remove these changes"
S
Shoaib Lari 已提交
255
						BACKOUT_COMMAND "if [ x$MASTER_HOSTNAME != x\`$HOSTNAME\` ];then $ECHO \"[FATAL]:-Not on original master host $MASTER_HOSTNAME, backout script exiting!\";exit 1;fi"
256 257 258
						$ECHO "$RM -f $BACKOUT_FILE" >> $BACKOUT_FILE
				fi
		fi
S
Shoaib Lari 已提交
259
		exit 1
260 261 262 263 264 265 266 267 268 269 270 271 272 273 274 275 276 277 278 279 280 281 282 283
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

ERROR_CHK () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	if [ $# -ne 3 ];then
		INITIAL_LEVEL=$DEBUG_LEVEL
		DEBUG_LEVEL=1
		LOG_MSG "[WARN]:-Incorrect # parameters supplied to $FUNCNAME"
		DEBUG_LEVEL=$INITIAL_LEVEL
		return;fi
	RETVAL=$1;shift
	MSG_TXT=$1;shift
	ACTION=$1 #1=issue warn, 2=fatal
	if [ $RETVAL -eq 0 ];then
		LOG_MSG "[INFO]:-Successfully completed $MSG_TXT"
	else
		if [ $ACTION -eq 1 ];then
			INITIAL_LEVEL=$DEBUG_LEVEL
			DEBUG_LEVEL=1
			LOG_MSG "[WARN]:-Issue with $MSG_TXT"
			DEBUG_LEVEL=$INITIAL_LEVEL
		else
			LOG_MSG "[INFO]:-End Function $FUNCNAME"
S
Shoaib Lari 已提交
284
			ERROR_EXIT "[FATAL]:-Failed to complete $MSG_TXT "
285
		fi
D
Daniel Gustafsson 已提交
286 287
	fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
288 289
}

290 291 292 293 294 295 296 297 298 299 300 301 302 303 304 305 306
RETRY () {
	RETVAL=$?
	if [[ "$CURRENT" =~ "ssh" ]]; then
		for i in 2 4 8; do
			sleep $i
			LOG_MSG "[WARN]:-Retrying command -- $CURRENT"
			eval "$CURRENT"
			if [ $? = 0 ]; then
				RETVAL=0
				# There seems to be no way of grabbing the return code of a
				# trap other than saving it to a variable
				return
			fi
		done
	fi
}

307
SED_PG_CONF () {
308 309 310 311 312 313
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	APPEND=0
	FILENAME=$1;shift
	SEARCH_TXT=$1;shift
	SUB_TXT="$1";shift
	KEEP_PREV=$1;shift
314 315
	SED_HOST=$1
	if [ x"" == x"$SED_HOST" ]; then
316 317 318 319 320 321 322 323 324 325
			if [ `$GREP -c "${SEARCH_TXT}[ ]*=" $FILENAME` -gt 1 ]; then
				LOG_MSG "[INFO]:-Found more than 1 instance of $SEARCH_TXT in $FILENAME, will append" 1
				APPEND=1
			fi
			if [ `$GREP -c "${SEARCH_TXT}[ ]*=" $FILENAME` -eq 0 ] || [ $APPEND -eq 1 ]; then
				$ECHO $SUB_TXT >> $FILENAME
				RETVAL=$?
				if [ $RETVAL -ne 0 ]; then
					LOG_MSG "[WARN]:-Failed to append line $SUB_TXT to $FILENAME" 1
				else
D
Daniel Gustafsson 已提交
326
					LOG_MSG "[INFO]:-Appended line $SUB_TXT to $FILENAME"
327 328
				fi
			else
329
				if [ $KEEP_PREV -eq 0 ];then
330
					$SED -i'.bak1' -e "s/${SEARCH_TXT}/${SUB_TXT} #${SEARCH_TXT}/" $FILENAME
331
				else
332
					$SED -i'.bak1' -e "s/${SEARCH_TXT}.*/${SUB_TXT}/" $FILENAME
333
				fi
334 335
				RETVAL=$?
				if [ $RETVAL -ne 0 ]; then
S
Shoaib Lari 已提交
336
					ERROR_EXIT "[FATAL]:-Failed to replace $SEARCH_TXT in $FILENAME"
337 338
				else
					LOG_MSG "[INFO]:-Replaced line in $FILENAME"
339
					$RM -f ${FILENAME}.bak1
340
				fi
341
				$SED -i'.bak2' -e "s/^#${SEARCH_TXT}/${SEARCH_TXT}/" $FILENAME
342 343
				RETVAL=$?
				if [ $RETVAL -ne 0 ]; then
S
Shoaib Lari 已提交
344
					ERROR_EXIT "[FATAL]:-Failed to replace #$SEARCH_TXT in $FILENAME"
345 346 347 348 349
				else
					LOG_MSG "[INFO]:-Replaced line in $FILENAME"
					$RM -f ${FILENAME}.bak2
				fi
			fi
350
	else
351 352 353 354 355 356 357 358
		# trap DEBUG will always be called first, when other traps are triggered.
		# We need to make sure that we save the current running command, so
		# that the RETRY function re-runs the command
		trap 'CURRENT=$BASH_COMMAND' DEBUG
		# Call out retry for commands that fail
		trap RETRY ERR
		RETVAL=0 # RETVAL gets modified in RETRY function whenever the trap is called

359 360
		if [ `$TRUSTED_SHELL $SED_HOST "$GREP -c \"${SEARCH_TXT}\" $FILENAME"` -gt 1 ]; then
			LOG_MSG "[INFO]:-Found more than 1 instance of $SEARCH_TXT in $FILENAME on $SED_HOST, will append" 1
361 362
			APPEND=1
		fi
363 364
		if [ `$TRUSTED_SHELL $SED_HOST "$GREP -c \"${SEARCH_TXT}\" $FILENAME"` -eq 0 ] || [ $APPEND -eq 1 ]; then
			$TRUSTED_SHELL $SED_HOST "$ECHO \"$SUB_TXT\" >> $FILENAME"
365
			if [ $RETVAL -ne 0 ]; then
S
Shoaib Lari 已提交
366
				ERROR_EXIT "[FATAL]:-Failed to append line $SUB_TXT to $FILENAME on $SED_HOST"
367
			else
368
				LOG_MSG "[INFO]:-Appended line $SUB_TXT to $FILENAME on $SED_HOST" 1
369 370 371
			fi
		else
			if [ $KEEP_PREV -eq 0 ];then
372
				SED_COMMAND="s/${SEARCH_TXT}/${SUB_TXT} #${SEARCH_TXT}/"
373
			else
374
				SED_COMMAND="s/${SEARCH_TXT}.*/${SUB_TXT}/"
375
			fi
376
			$TRUSTED_SHELL $SED_HOST sed -i'.bak1' -f /dev/stdin "$FILENAME" <<< "$SED_COMMAND" > /dev/null 2>&1
377
			if [ $RETVAL -ne 0 ]; then
S
Shoaib Lari 已提交
378
				ERROR_EXIT "[FATAL]:-Failed to insert $SUB_TXT in $FILENAME on $SED_HOST"
379
			else
380
				LOG_MSG "[INFO]:-Replaced line in $FILENAME on $SED_HOST"
381 382
				$TRUSTED_SHELL $SED_HOST "$RM -f ${FILENAME}.bak1" > /dev/null 2>&1
			fi
383 384 385

			SED_COMMAND="s/^#${SEARCH_TXT}/${SEARCH_TXT}/"
			$TRUSTED_SHELL $SED_HOST sed -i'.bak2' -f /dev/stdin "$FILENAME" <<< "$SED_COMMAND" > /dev/null 2>&1
386
			if [ $RETVAL -ne 0 ]; then
S
Shoaib Lari 已提交
387
				ERROR_EXIT "[FATAL]:-Failed to substitute #${SEARCH_TXT} in $FILENAME on $SED_HOST"
388 389 390
			else
				LOG_MSG "[INFO]:-Replaced line in $FILENAME on $SED_HOST"
				$TRUSTED_SHELL $SED_HOST "$RM -f ${FILENAME}.bak2" > /dev/null 2>&1
391 392
			fi
		fi
393 394

		trap - ERR DEBUG # Disable trap
395 396 397 398 399 400 401 402
	fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

POSTGRES_PORT_CHK () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	GET_PG_PID_ACTIVE $1 $2
	if [ $PID -ne 0 ];then
S
Shoaib Lari 已提交
403
		ERROR_EXIT "[FATAL]:-Host $2 has an active database process on port = $1"
404 405 406 407 408 409 410 411 412 413 414
	fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

CREATE_SPREAD_MIRROR_ARRAY () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	((MAX_ARRAY=${#QE_PRIMARY_ARRAY[@]}-1))

	# Current host and subnet we are working on
	CURRENT_HOST=0
	CURRENT_SUBNET=0
D
Daniel Gustafsson 已提交
415

416 417 418
	# Destination host and subnet
	DEST_HOST=0
	DEST_SUBNET=0
D
Daniel Gustafsson 已提交
419

420 421 422 423 424
	if [ x"$NUM_MHOST_NODE" != x"" ] && [ $NUM_MHOST_NODE -gt 0 ] ; then
		((DIRS_PER_SUBNET=$NUM_DATADIR/$NUM_MHOST_NODE))
	else
		DIRS_PER_SUBNET=$NUM_DATADIR
	fi
D
Daniel Gustafsson 已提交
425

426 427
	((MAX_SUBNET=$NUM_DATADIR/$DIRS_PER_SUBNET))
	((MAX_HOST=${#QE_PRIMARY_ARRAY[@]}/$NUM_DATADIR))
D
Daniel Gustafsson 已提交
428

429 430 431 432 433
	SEGS_PROCESSED=0
	SEGS_PROCESSED_HOST=0


	# The following is heavily dependent on sort order of primary array.  This sort
D
Daniel Gustafsson 已提交
434 435
	# order will be affected by hostnames so something non-standard will cause
	# strange behaviour.  This isn't new (just recording this fact for future generations)
436
	# and can be worked around with a mapping file to gpinitsystem (-I option).
D
Daniel Gustafsson 已提交
437
	# The right way to do this would require us to connect to remote hosts, determine
438 439 440 441 442 443 444 445 446 447 448 449 450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468
	# what subnet we are on for that hostname and then build the array that way.  We *will*
	# do this once this is in python (or anything other than BASH)
	LOG_MSG "[INFO]:-Building spread mirror array type $MULTI_TXT, please wait..." 1
	for QE_LINE in ${QE_PRIMARY_ARRAY[@]}
	do
		if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $NOLINE_ECHO ".\c";fi

		if [ $(($SEGS_PROCESSED%$NUM_DATADIR)) -eq 0 ] ; then
			# A new host group is starting
			if [ $SEGS_PROCESSED -ne 0 ] ; then ((CURRENT_HOST=$CURRENT_HOST+1)); fi
			# Start the mirroring on the next host
			((DEST_HOST=$CURRENT_HOST+1))
			# Always subnet "0" to start
			CURRENT_SUBNET=0
			DEST_SUBNET=1
			# Make sure we loop back when needed
			if [ $DEST_HOST -ge $MAX_HOST ] ; then DEST_HOST=0; fi
			SEGS_PROCESSED_HOST=0
		else
			# Continue with current host
			# move dest host to the next one (This is spread mirroring)
			((DEST_HOST=$DEST_HOST+1))
			# Make sure we look back when needed
			if [ $DEST_HOST -ge $MAX_HOST ] ; then DEST_HOST=0; fi
			# Get what subnet we are on, we may have moved to next
			((CURRENT_SUBNET=($SEGS_PROCESSED_HOST+1)/$DIRS_PER_SUBNET))
			((DEST_SUBNET=$CURRENT_SUBNET+1))
			# Handle looping over
			if [ $DEST_SUBNET -ge $MAX_SUBNET ] ; then DEST_SUBNET=0; fi
			# Increment the number of segments we've processed for this host
			((SEGS_PROCESSED_HOST=$SEGS_PROCESSED_HOST+1))
D
Daniel Gustafsson 已提交
469 470
		fi

471 472 473
        # Handle the case where it's a single hostname (thus a single subnet)
		# This case will mainly be for QA testing
		if [ $NUM_DATADIR -eq $DIRS_PER_SUBNET ] ; then DEST_SUBNET=0; fi
D
Daniel Gustafsson 已提交
474

475 476 477 478 479 480 481 482 483 484 485 486
		# Handle possible loop
		if [ $DEST_SUBNET -ge $MAX_SUBNET ] ; then DEST_SUBNET=0; fi

		# Calculate the index based on host and subnet number
		((PRIM_SEG_INDEX=($DEST_HOST*$NUM_DATADIR)+($DEST_SUBNET*$DIRS_PER_SUBNET)))

		QE_M_NAME=`$ECHO ${QE_PRIMARY_ARRAY[$PRIM_SEG_INDEX]}|$AWK -F"~" '{print $1}'`
		GP_M_DIR=${MIRROR_DATA_DIRECTORY[$SEGS_PROCESSED%$NUM_DATADIR]}
		P_PORT=`$ECHO $QE_LINE|$AWK -F"~" '{print $2}'`
		((GP_M_PORT=$P_PORT+$MIRROR_OFFSET))
		M_CONTENT=`$ECHO $QE_LINE|$AWK -F"~" '{print $5}'`
		M_SEG=`$ECHO $QE_LINE|$AWK -F"~" '{print $3}'|$AWK -F"/" '{print $NF}'`
487
		QE_MIRROR_ARRAY=(${QE_MIRROR_ARRAY[@]} ${QE_M_NAME}~${GP_M_PORT}~${GP_M_DIR}/${M_SEG}~${DBID_COUNT}~${M_CONTENT})
488 489 490 491 492 493 494 495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513
		POSTGRES_PORT_CHK $GP_M_PORT $QE_M_NAME
		((DBID_COUNT=$DBID_COUNT+1))
		((SEGS_PROCESSED=$SEGS_PROCESSED+1))
	done
	if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $ECHO;fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

CREATE_GROUP_MIRROR_ARRAY () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	LOG_MSG "[INFO]:-Building group mirror array type $MULTI_TXT, please wait..." 1
	PRI_HOST_COUNT=`$ECHO ${QE_PRIMARY_ARRAY[@]}|$TR ' ' '\n'|$AWK -F"~" '{print $1}'|$SORT -u|$WC -l`
	if [ $MULTI_HOME -eq 1 ] && [ $REMOTE_HOST_COUNT -eq 1 ];then
		PRI_HOST_COUNT=1
	fi

	if [ x"$NUM_MHOST_NODE" != x"" ] && [ $NUM_MHOST_NODE -gt 0 ] ; then
		((DIRS_PER_SUBNET=$NUM_DATADIR/$NUM_MHOST_NODE))
	else
		DIRS_PER_SUBNET=$NUM_DATADIR
	fi
	((MAX_SUBNET=$NUM_DATADIR/$DIRS_PER_SUBNET))
	((MAX_HOST=${#QE_PRIMARY_ARRAY[@]}/$NUM_DATADIR))

	# Current host we are working on
	CURRENT_HOST=0
D
Daniel Gustafsson 已提交
514

515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531
	# Destination host and subnet
	DEST_HOST=0
	DEST_SUBNET=0

	PRIMARY_ARRAY_LENGTH=${#QE_PRIMARY_ARRAY[@]}
	PRIMARY_INDEX=0

	for QE_LINE in ${QE_PRIMARY_ARRAY[@]}
	do
		if [ $(($PRIMARY_INDEX%$NUM_DATADIR)) -eq 0 ] ; then
			if [ $PRIMARY_INDEX -ne 0 ] ; then ((CURRENT_HOST=$CURRENT_HOST+1)); fi
			((DEST_HOST=$CURRENT_HOST+1))
			if [ $DEST_HOST -ge $MAX_HOST ] ; then DEST_HOST=0; fi
			DEST_SUBNET=1
		else
			if [ $(($PRIMARY_INDEX%$DIRS_PER_SUBNET)) -eq 0 ] ; then
				((DEST_SUBNET=$DEST_SUBNET+1))
D
Daniel Gustafsson 已提交
532
			fi
533 534 535
		fi

		# Handle possible loop
D
Daniel Gustafsson 已提交
536
		if [ $DEST_SUBNET -ge $MAX_SUBNET ] ; then DEST_SUBNET=0; fi
537 538 539 540 541 542 543 544 545 546 547

		((MIRROR_INDEX=($DEST_HOST*$NUM_DATADIR)+($DEST_SUBNET*$DIRS_PER_SUBNET)))

		if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $NOLINE_ECHO ".\c";fi

		QE_M_NAME=`$ECHO ${QE_PRIMARY_ARRAY[$MIRROR_INDEX]}|$AWK -F"~" '{print $1}'`
		GP_M_DIR=${MIRROR_DATA_DIRECTORY[$PRIMARY_INDEX%$NUM_DATADIR]}/`$ECHO $QE_LINE|$AWK -F"~" '{print $3}'|$AWK -F"/" '{print $NF}'`

		M_CONTENT=`$ECHO $QE_LINE|$AWK -F"~" '{print $5}'`
		P_PORT=`$ECHO $QE_LINE|$AWK -F"~" '{print $2}'`
		GP_M_PORT=$(($P_PORT+$MIRROR_OFFSET))
D
Daniel Gustafsson 已提交
548

549
		QE_MIRROR_ARRAY=(${QE_MIRROR_ARRAY[@]} ${QE_M_NAME}~${GP_M_PORT}~${GP_M_DIR}~${DBID_COUNT}~${M_CONTENT})
550 551 552 553 554 555 556 557 558 559
		POSTGRES_PORT_CHK $GP_M_PORT $QE_M_NAME

		DBID_COUNT=$(($DBID_COUNT+1))
	    PRIMARY_INDEX=$((PRIMARY_INDEX+1))
	done
	if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $ECHO;fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

GET_REPLY () {
560 561
	$ECHO -e "\n$1 Yy|Nn (default=N):"
	$ECHO -n "> "
562 563 564 565 566 567 568 569 570
	read REPLY
	if [ -z $REPLY ]; then
		LOG_MSG "[WARN]:-User abort requested, Script Exits!" 1
		exit 1
	fi
	if [ $REPLY != Y ] && [ $REPLY != y ]; then
		LOG_MSG "[WARN]:-User abort requested, Script Exits!" 1
		exit 1
	fi
D
Daniel Gustafsson 已提交
571 572
}

573 574 575 576 577 578 579 580 581 582 583 584 585 586 587 588 589 590 591 592 593 594 595
CHK_FILE () {
		LOG_MSG "[INFO]:-Start Function $FUNCNAME"
		FILENAME=$1
		FILE_HOST=$2
		if [ x"" == x"$FILE_HOST" ];then
			LOG_MSG "[INFO]:-Checking file $FILENAME"
			if [ ! -s $FILENAME ] || [ ! -r $FILENAME ]
					then
					EXISTS=1
			else
					EXISTS=0
			fi
		else
			EXISTS=`$TRUSTED_SHELL $FILE_HOST "if [ ! -s $FILENAME ] || [ ! -r $FILENAME ];then $ECHO 1;else $ECHO 0;fi"`
			RETVAL=$?
			if [ $RETVAL -ne 0 ];then
				LOG_MSG "[WARN]:-Failed to obtain details of $FILENAME on $FILE_HOST"
				EXISTS=1
			fi
		fi
		LOG_MSG "[INFO]:-End Function $FUNCNAME"
}
CHK_DIR () {
596 597 598 599 600 601
		# this function might be called very early, before logfiles are initialized
		if [ x"" == x"$3" ];then
			LOG_MSG "[INFO]:-Start Function $FUNCNAME"
		fi
		DIR_NAME=$1
		DIR_HOST=$2
602 603 604 605 606 607 608 609 610 611
		if [ x"" == x"$DIR_HOST" ];then
			EXISTS=`if [ -d $DIR_NAME ];then $ECHO 0;else $ECHO 1;fi`
		else
			EXISTS=`$TRUSTED_SHELL $DIR_HOST "if [ -d $DIR_NAME ];then $ECHO 0;else $ECHO 1;fi"`
			RETVAL=$?
			if [ $RETVAL -ne 0 ];then
			LOG_MSG "[WARN]:-Failed to obtain details of $DIR_NAME on $DIR_HOST" 1
			EXISTS=1
			fi
		fi
612 613 614
		if [ x"" == x"$3" ];then
			LOG_MSG "[INFO]:-End Function $FUNCNAME"
		fi
615 616 617 618 619 620
}

GET_MASTER_PORT () {
		LOG_MSG "[INFO]:-Start Function $FUNCNAME"
		MASTER_DATA_DIRECTORY=$1
		if [ x"" == x"$MASTER_DATA_DIRECTORY" ];then
S
Shoaib Lari 已提交
621
			ERROR_EXIT "[FATAL]:-MASTER_DATA_DIRECTORY variable not set";fi
622
		if [ ! -d $MASTER_DATA_DIRECTORY ]; then
S
Shoaib Lari 已提交
623
				ERROR_EXIT "[FATAL]:-No $MASTER_DATA_DIRECTORY directory"
624 625 626 627 628 629 630 631 632 633 634 635 636
		fi
		if [ -r $MASTER_DATA_DIRECTORY/$PG_CONF ];then
			MASTER_PORT=`$AWK 'split($0,a,"#")>0 && split(a[1],b,"=")>1 {print b[1] " " b[2]}' $MASTER_DATA_DIRECTORY/$PG_CONF | $AWK '$1=="port" {print $2}' | $TAIL -1`
			if [ x"" == x"$MASTER_PORT" ] ; then
                #look for include files
                for INC_FILE in `$AWK '/^[ ]*include /{print $2}' $MASTER_DATA_DIRECTORY/$PG_CONF | $TR -d "'\""` ; do
                    if [[ $INC_FILE == /* ]] ; then
                        GET_MASTER_PORT_RECUR "$INC_FILE" 1
                    else
                        GET_MASTER_PORT_RECUR "$MASTER_DATA_DIRECTORY/$INC_FILE" 1
                    fi
                done
                if [ x"" == x"$MASTER_PORT" ] ; then
S
Shoaib Lari 已提交
637
			        ERROR_EXIT "[FATAL]:-Failed to obtain master port number from $MASTER_DATA_DIRECTORY/$PG_CONF"
638 639 640
                fi
			fi
		else
S
Shoaib Lari 已提交
641
			ERROR_EXIT "[FATAL]:-Do not have read access to $MASTER_DATA_DIRECTORY/$PG_CONF"
642 643 644 645 646 647 648 649 650 651 652 653 654 655 656 657 658 659 660 661 662 663 664 665 666 667 668 669 670 671 672 673 674 675 676 677 678 679 680 681 682 683 684
		fi
		LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

GET_MASTER_PORT_RECUR () {
    INCLUDED_FILE=$1
    RECUR=$2
    if [ $RECUR -le 10 ] ; then
        MASTER_PORT=`$AWK 'split($0,a,"#")>0 && split(a[1],b,"=")>1 {print b[1] " " b[2]}' $INCLUDED_FILE | $AWK '$1=="port" {print $2}' | $TAIL -1`
        if [ x"" == x"$MASTER_PORT" ] ; then
            #look for include files
            let CURR_DEPTH=$RECUR+1
            for INC_FILE in `$AWK '/^[ ]*include /{print $2}' $INC_FILE | $TR -d "'\""` ; do
                if [[ $INC_FILE == /* ]] ; then
                    GET_MASTER_PORT_RECUR "$INC_FILE" $CURR_DEPTH
                else
                    GET_MASTER_PORT_RECUR "$MASTER_DATA_DIRECTORY/$INC_FILE" $CURR_DEPTH
                fi
                if [ x"" != x"$MASTER_PORT" ] ; then
                    break
                fi
            done
        fi
    else
        ERROR_EXIT "[FATAL]:-Could not open configuration file \"$INCLUDED_FILE\": maximum nesting depth exceeded"
    fi
}

GET_CIDRADDR () {
    # MPP-15889
    # assuming argument is an ip address, return the address
    # with a /32 or /128 cidr suffix based on whether or not the
    # address contains a :

    if [ `echo $1 | grep -c :` -gt 0 ]; then
	echo $1/128
    else
	echo $1/32
    fi
}

BUILD_MASTER_PG_HBA_FILE () {
        LOG_MSG "[INFO]:-Start Function $FUNCNAME"
S
Shoaib Lari 已提交
685
	if [ $# -eq 0 ];then ERROR_EXIT "[FATAL]:-Passed zero parameters, expected at least 2";fi
686
	GP_DIR=$1
687
	HBA_HOSTNAMES=${2:-0}
688 689 690 691 692 693 694
        LOG_MSG "[INFO]:-Clearing values in Master $PG_HBA"
        $GREP "^#" ${GP_DIR}/$PG_HBA > $TMP_PG_HBA
        $MV $TMP_PG_HBA ${GP_DIR}/$PG_HBA
        LOG_MSG "[INFO]:-Setting local access"
        $ECHO "local    all         $USER_NAME         $PG_METHOD" >> ${GP_DIR}/$PG_HBA
        #$ECHO "local    all         all                $PG_METHOD" >> ${GP_DIR}/$PG_HBA
        LOG_MSG "[INFO]:-Setting local host access"
695
        if [ $HBA_HOSTNAMES -eq 0 ];then
696 697 698 699 700 701 702 703 704 705 706 707 708 709 710 711 712 713 714 715 716 717 718 719 720 721 722 723
            $ECHO "host     all         $USER_NAME         127.0.0.1/28    trust" >> ${GP_DIR}/$PG_HBA

            for ADDR in "${MASTER_IP_ADDRESS_ALL[@]}"
            do
                # MPP-15889
                CIDRADDR=$(GET_CIDRADDR $ADDR)
                $ECHO "host     all         $USER_NAME         $CIDRADDR       trust" >> ${GP_DIR}/$PG_HBA

            done
            for ADDR in "${STANDBY_IP_ADDRESS_ALL[@]}"
            do
                # MPP-15889
                CIDRADDR=$(GET_CIDRADDR $ADDR)
                $ECHO "host     all         $USER_NAME         $CIDRADDR       trust" >> ${GP_DIR}/$PG_HBA
            done

            # Add all local IPV6 addresses
            for ADDR in "${MASTER_IPV6_LOCAL_ADDRESS_ALL[@]}"
            do
                # MPP-15889
                CIDRADDR=$(GET_CIDRADDR $ADDR)
                $ECHO "host     all         $USER_NAME         $CIDRADDR       trust" >> ${GP_DIR}/$PG_HBA
            done
        else
            $ECHO "host     all         $USER_NAME         localhost    trust" >> ${GP_DIR}/$PG_HBA
            $ECHO "host     all         $USER_NAME         $MASTER_HOSTNAME       trust" >> ${GP_DIR}/$PG_HBA
        fi

724 725 726 727 728 729 730 731

        # Add replication config
        $ECHO "local    replication $USER_NAME         $PG_METHOD" >> ${GP_DIR}/$PG_HBA
        $ECHO "host     replication $USER_NAME         samenet       trust" >> ${GP_DIR}/$PG_HBA
        LOG_MSG "[INFO]:-Complete Master $PG_HBA configuration"
        LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

732 733
BUILD_GPSSH_CONF () {
        LOG_MSG "[INFO]:-Start Function $FUNCNAME"
S
Shoaib Lari 已提交
734
        if [ $# -eq 0 ];then ERROR_EXIT "[FATAL]:-Passed zero parameters, expected at least 1";fi
735 736 737 738 739 740 741 742 743 744 745 746 747 748
        GP_DIR=$1
        $CAT <<_EOF_ >> $GP_DIR/gpssh.conf
[gpssh]
# delaybeforesend specifies the time in seconds to wait at the
# beginning of an ssh interaction before doing anything.
# Increasing this value can have a big runtime impact at the
# beginning of gpssh.
delaybeforesend = 0.05

# prompt_validation_timeout specifies a timeout multiplier that
# will be used in validating the ssh prompt. Increasing this
# value will have a small runtime impact at the beginning of
# gpssh.
prompt_validation_timeout = 1.0
749 750 751 752 753 754 755 756 757 758

# sync_retries specifies how many times to try the pxssh
# connection verification.
# Setting this value to 1 means gpssh will immediately pass
# along pxssh's best effort.
# Increasing this value will allow for slow network connections,
# cpu load, or other slowness on the segment host, but will
# also delay feedback when a connection cannot be established
# for other reasons
sync_retries = 3
759 760 761 762
_EOF_
        LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

763 764 765 766 767 768
BUILD_PERFMON() {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	GP_DIR=$1
	$MKDIR -p $GP_DIR/gpperfmon/conf $GP_DIR/gpperfmon/logs $GP_DIR/gpperfmon/data
	$CAT <<_EOF_ >> $GP_DIR/gpperfmon/conf/gpperfmon.conf
[GPMMON]
D
Daniel Gustafsson 已提交
769 770
# quantum specifies the time in seconds between updates from
# performance monitor agents on all segments. Valid values
771 772 773
# are 10, 15, 20, 30, or 60
quantum = 15

D
Daniel Gustafsson 已提交
774 775 776 777
# min_query_time specifies the minimum query run time
# in seconds for statistics collection. The monitor logs all
# queries that run longer than this value in the queries_history
# table. For queries with shorter run times, no historical
778 779 780 781
# data is collected.
min_query_time = 20

# This should be a percentage between 0 and 100 and should be
782
# less than the error_disk_space_percentage.  If a filesystem's
D
Daniel Gustafsson 已提交
783 784 785
# disk space used percentage equals or exceeds this value a
# warning will be logged and a warning email/snmp trap may be
# sent.  If this configuration is set to 0 or not specified, no
786 787 788
# warnings are sent.
#warning_disk_space_percentage = 80

D
Daniel Gustafsson 已提交
789 790
# This should be a percentage between 0 and 100 and should be
# greater than the warning_disk_space_percentage. If a
791
# filesystem's disk space used percentage equals or exceeds
D
Daniel Gustafsson 已提交
792 793
# this value an error will be logged and a error email/snmp
# trap may be sent.  If this configuration is set to 0 or not
794 795 796
# specified, no errors are sent.
#error_disk_space_percentage = 90

D
Daniel Gustafsson 已提交
797 798 799 800
#This is the interval in minutes that limits the number of
#error/warning messages that are sent. The minimum value for
#this configuration is 1.  Setting this to 0 or not specifying
#this configuration results in it getting set to the minimum.
801 802
disk_space_interval = 60

D
Daniel Gustafsson 已提交
803
#This is the maximum number of error/warning messages that
804
#will be sent in the disk_space_interval.  The maximum value
D
Daniel Gustafsson 已提交
805 806 807
#for this configuration is 50.  The minimum value for this
#configuration is 1.  Setting this configuration to greater
#than 50 or not specifying this configuration results in it
808 809 810
#getting set to the maximum.
max_disk_space_messages_per_interval = 10

D
Daniel Gustafsson 已提交
811
# The number of partitions for statistics data in month
812 813 814 815 816 817 818 819 820 821 822 823 824 825 826 827 828 829 830 831 832 833 834 835 836 837 838 839 840 841 842 843 844 845 846 847 848 849 850 851 852 853 854 855 856 857 858 859 860 861 862 863 864 865 866 867 868 869 870 871 872 873 874 875
# will be retained. Older partitions will be dropped.
#partition_age = 6


log_location = $GP_DIR/gpperfmon/logs
_EOF_
}

GET_PG_PID_ACTIVE () {
		LOG_MSG "[INFO]:-Start Function $FUNCNAME"
		# Expects port number and hostname for remote checking
		PORT=$1;shift
		HOST=$1
		PG_LOCK_FILE="/tmp/.s.PGSQL.${PORT}.lock"
		PG_LOCK_NETSTAT=""
		if [ x"" == x"$HOST" ];then
			#See if we have a netstat entry for this local host
			PORT_ARRAY=(`$NETSTAT -an 2>/dev/null |$GREP ".s.PGSQL.${PORT}"|$AWK '{print $NF}'|$AWK -F"." '{print $NF}'|$SORT -u`)
			for P_CHK in ${PORT_ARRAY[@]}
			do
				if [ $P_CHK -eq $PORT ];then  PG_LOCK_NETSTAT=$PORT;fi
			done
			#PG_LOCK_NETSTAT=`$NETSTAT -an 2>/dev/null |$GREP ".s.PGSQL.${PORT}"|$AWK '{print $NF}'|$HEAD -1`
			#See if we have a lock file in /tmp
			if [ -f ${PG_LOCK_FILE} ];then
				PG_LOCK_TMP=1
			else
				PG_LOCK_TMP=0
			fi
			if [ x"" == x"$PG_LOCK_NETSTAT" ] && [ $PG_LOCK_TMP -eq 0 ];then
				PID=0
				LOG_MSG "[INFO]:-No socket connection or lock file in /tmp found for port=${PORT}"
			else
				#Now check the failure combinations
				if [ $PG_LOCK_TMP -eq 0 ] && [ x"" != x"$PG_LOCK_NETSTAT" ];then
				#Have a process but no lock file
					LOG_MSG "[WARN]:-No lock file $PG_LOCK_FILE but process running on port $PORT" 1
					PID=1
				fi
				if [ $PG_LOCK_TMP -eq 1 ] && [ x"" == x"$PG_LOCK_NETSTAT" ];then
				#Have a lock file but no process
					if [ -r ${PG_LOCK_FILE} ];then
						PID=`$CAT ${PG_LOCK_FILE}|$HEAD -1|$AWK '{print $1}'`
					else
						LOG_MSG "[WARN]:-Unable to access ${PG_LOCK_FILE}" 1
						PID=1
					fi
					LOG_MSG "[WARN]:-Have lock file $PG_LOCK_FILE but no process running on port $PORT" 1
				fi
				if [ $PG_LOCK_TMP -eq 1 ] && [ x"" != x"$PG_LOCK_NETSTAT" ];then
				#Have both a lock file and a netstat process
					if [ -r ${PG_LOCK_FILE} ];then
						PID=`$CAT ${PG_LOCK_FILE}|$HEAD -1|$AWK '{print $1}'`
					else
						LOG_MSG "[WARN]:-Unable to access ${PG_LOCK_FILE}" 1
						PID=1
					fi
					LOG_MSG "[INFO]:-Have lock file $PG_LOCK_FILE and a process running on port $PORT"
				fi
			fi
		else
			PING_HOST $HOST 1
			if [ $RETVAL -ne 0 ];then
				PID=0
D
Daniel Gustafsson 已提交
876
			else
877 878 879 880 881 882
				PORT_ARRAY=(`$TRUSTED_SHELL $HOST "$NETSTAT -an 2>/dev/null |$GREP ".s.PGSQL.${PORT}" 2>/dev/null"|$AWK '{print $NF}'|$AWK -F"." '{print $NF}'|$SORT -u`)
				for P_CHK in ${PORT_ARRAY[@]}
				do
					if [ $P_CHK -eq $PORT ];then  PG_LOCK_NETSTAT=$PORT;fi
				done
				#PG_LOCK_NETSTAT=`$TRUSTED_SHELL $HOST "$NETSTAT -an 2>/dev/null |$GREP ".s.PGSQL.${PORT}" 2>/dev/null"|$AWK '{print $NF}'|$HEAD -1`
D
Daniel Gustafsson 已提交
883
				PG_LOCK_TMP=`$TRUSTED_SHELL $HOST "ls ${PG_LOCK_FILE} 2>/dev/null"|$WC -l`
884 885 886 887 888 889 890 891 892 893 894 895 896 897 898 899 900 901 902 903 904 905 906 907 908 909 910
				if [ x"" == x"$PG_LOCK_NETSTAT" ] && [ $PG_LOCK_TMP -eq 0 ];then
					PID=0
					LOG_MSG "[INFO]:-No socket connection or lock file $PG_LOCK_FILE found for port=${PORT}"
				else
				#Now check the failure combinations
					if [ $PG_LOCK_TMP -eq 0 ] && [ x"" != x"$PG_LOCK_NETSTAT" ];then
					#Have a process but no lock file
						LOG_MSG "[WARN]:-No lock file $PG_LOCK_FILE but process running on port $PORT on $HOST" 1
						PID=1
					fi
					if [ $PG_LOCK_TMP -eq 1 ] && [ x"" == x"$PG_LOCK_NETSTAT" ];then
					#Have a lock file but no process
						CAN_READ=`$TRUSTED_SHELL $HOST "if [ -r ${PG_LOCK_FILE} ];then echo 1;else echo 0;fi"`
						if [ $CAN_READ -eq 1 ];then
							PID=`$TRUSTED_SHELL $HOST "$CAT ${PG_LOCK_FILE}|$HEAD -1 2>/dev/null"|$AWK '{print $1}'`
						else
							LOG_MSG "[WARN]:-Unable to access ${PG_LOCK_FILE} on $HOST" 1
						fi
						LOG_MSG "[WARN]:-Have lock file $PG_LOCK_FILE but no process running on port $PORT on $HOST" 1
						PID=1
					fi
					if [ $PG_LOCK_TMP -eq 1 ] && [ x"" != x"$PG_LOCK_NETSTAT" ];then
					#Have both a lock file and a netstat process
						CAN_READ=`$TRUSTED_SHELL $HOST "if [ -r ${PG_LOCK_FILE} ];then echo 1;else echo 0;fi"`
						if [ $CAN_READ -eq 1 ];then
							PID=`$TRUSTED_SHELL $HOST "$CAT ${PG_LOCK_FILE}|$HEAD -1 2>/dev/null"|$AWK '{print $1}'`
						else
D
Daniel Gustafsson 已提交
911
							LOG_MSG "[WARN]:-Unable to access ${PG_LOCK_FILE} on $HOST" 1
912 913
						fi
						LOG_MSG "[INFO]:-Have lock file $PG_LOCK_FILE and a process running on port $PORT on $HOST"
D
Daniel Gustafsson 已提交
914
					fi
915 916 917 918 919 920 921 922 923 924 925 926 927 928 929 930 931 932 933 934 935 936 937 938 939 940 941 942 943 944 945 946 947 948 949 950 951 952 953 954 955
				fi
			fi
		fi
		LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

RUN_COMMAND_REMOTE () {
		LOG_MSG "[INFO]:-Start Function $FUNCNAME"
		HOST=$1
		COMMAND=$2
		LOG_MSG "[INFO]:-Commencing remote $TRUSTED_SHELL $HOST $COMMAND"
		$TRUSTED_SHELL $HOST $COMMAND >> $LOG_FILE 2>&1
		RETVAL=$?
		if [ $RETVAL -ne 0 ]; then
			LOG_MSG "[FATAL]:- Command $COMMAND on $HOST failed with error status $RETVAL" 2
		else
			LOG_MSG "[INFO]:-Completed $TRUSTED_SHELL $HOST $COMMAND"
		fi
		LOG_MSG "[INFO]:-End Function $FUNCNAME"
		return $RETVAL
}

BACKOUT_COMMAND () {
		LOG_MSG "[INFO]:-Start Function $FUNCNAME"
		COMMAND=$1
		if [ ! -f $BACKOUT_FILE ]; then
				$ECHO $COMMAND > $BACKOUT_FILE
		else
				$CAT $BACKOUT_FILE > /tmp/backout_file.tmp.$$
				$ECHO $COMMAND > $BACKOUT_FILE
				$CAT /tmp/backout_file.tmp.$$ >> $BACKOUT_FILE
				$RM -f /tmp/backout_file.tmp.$$
		fi
		LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

PING_HOST () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	TARGET_HOST=$1;shift
	PING_EXIT=$1
	if [ x"" == x"$PING_EXIT" ];then PING_EXIT=0;fi
956
	OUTPUT=""
957 958
	case $OS_TYPE in
		darwin )
959
			OUTPUT=$($PING $PING_TIME $TARGET_HOST 2>&1 || $PING6 $PING_TIME $TARGET_HOST 2>&1)
960
                        ;;
961
		linux )
962
			OUTPUT=$($PING $TARGET_HOST $PING_TIME 2>&1 || $PING6 $TARGET_HOST $PING_TIME 2>&1)
963
                        ;;
964
		openbsd )
965
			OUTPUT=$($PING $PING_TIME $TARGET_HOST 2>&1 || $PING6 $PING_TIME $TARGET_HOST 2>&1)
966
                        ;;
967
		* )
968
			OUTPUT=$($PING $TARGET_HOST $PING_TIME 2>&1)
969 970 971
	esac
	RETVAL=$?
	case $RETVAL in
D
Daniel Gustafsson 已提交
972
		0) LOG_MSG "[INFO]:-$TARGET_HOST contact established"
973
                   ;;
974
		1) if [ $PING_EXIT -eq 0 ];then
S
Shoaib Lari 已提交
975
			ERROR_EXIT "[FATAL]:-Unable to contact $TARGET_HOST: $OUTPUT"
976
		   else
977
			LOG_MSG "[WARN]:-Unable to contact $TARGET_HOST: $OUTPUT" 1
D
Daniel Gustafsson 已提交
978
		   fi
979
                   ;;
980
		2) if [ $PING_EXIT -eq 0 ];then
S
Shoaib Lari 已提交
981
			ERROR_EXIT "[FATAL]:-Unknown host $TARGET_HOST: $OUTPUT"
982
		   else
983 984 985 986
			LOG_MSG "[WARN]:-Unknown host $TARGET_HOST: $OUTPUT" 1
		   fi
                   ;;
		*) if [ $PING_EXIT -eq 0 ];then
S
Shoaib Lari 已提交
987
			ERROR_EXIT "[FATAL]:-Cannot ping host $TARGET_HOST: $OUTPUT"
988 989
		   else
			LOG_MSG "[WARN]:-Cannot ping host $TARGET_HOST: $OUTPUT" 1
D
Daniel Gustafsson 已提交
990
		   fi
991
                   ;;
992 993 994 995 996 997 998 999 1000 1001 1002 1003 1004 1005 1006 1007 1008 1009 1010 1011
	esac
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
	return $RETVAL
}

PARALLEL_SETUP () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	PARALLEL_STATUS_FILE=$1
	$TOUCH $PARALLEL_STATUS_FILE
	export PARALLEL_STATUS_FILE=$PARALLEL_STATUS_FILE
	LOG_MSG "[INFO]:-Spawning parallel processes    batch [1], please wait..." 1
	BATCH_COUNT=0
	INST_COUNT=0
	BATCH_DONE=1
	BATCH_TOTAL=0
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

PARALLEL_COUNT () {
        LOG_MSG "[INFO]:-Start Function $FUNCNAME"
S
Shoaib Lari 已提交
1012
	if [ $# -ne 2 ];then ERROR_EXIT "[FATAL]:-Incorrect number of parameters passed to $FUNCNAME";fi
1013 1014 1015 1016 1017 1018 1019 1020 1021 1022 1023 1024 1025 1026 1027 1028 1029 1030 1031 1032 1033 1034 1035 1036 1037 1038 1039 1040 1041 1042
	BATCH_LIMIT=$1
	BATCH_DEFAULT=$2
	((INST_COUNT=$INST_COUNT+1))
	((BATCH_COUNT=$BATCH_COUNT+1))
	((BATCH_TOTAL=$BATCH_TOTAL+1))
	if [ $BATCH_COUNT -eq $BATCH_DEFAULT ] || [ $BATCH_LIMIT -eq $BATCH_TOTAL ];then
		if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $ECHO;fi
		PARALLEL_WAIT
		((BATCH_DONE=$BATCH_DONE+1))
		BATCH_COUNT=0
		if [ $BATCH_LIMIT -ne $BATCH_TOTAL ];then
			LOG_MSG "[INFO]:-Spawning parallel processes    batch [$BATCH_DONE], please wait..." 1
		fi
	fi
        LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

PARALLEL_WAIT () {
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
	LOG_MSG "[INFO]:-Waiting for parallel processes batch [$BATCH_DONE], please wait..." 1
	SLEEP_COUNT=0
	while [ `$WC -l $PARALLEL_STATUS_FILE|$AWK '{print $1}'` -ne $INST_COUNT ]
	do
		if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $NOLINE_ECHO ".\c";fi
		$SLEEP 1
		((SLEEP_COUNT=$SLEEP_COUNT+1))
		if [ $WAIT_LIMIT -lt $SLEEP_COUNT ];then
			if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $NOLINE_ECHO ".\c";fi
			LOG_MSG "[FATAL]:-Failed to process this batch of segments within $WAIT_LIMIT seconds" 1
			LOG_MSG "[INFO]:-Review contents of $LOG_FILE" 1
S
Shoaib Lari 已提交
1043
			ERROR_EXIT "[FATAL]:-Process timeout failure"
1044 1045 1046 1047 1048 1049 1050
		fi
	done
	if [ $DEBUG_LEVEL -eq 0 ] && [ x"" != x"$VERBOSE" ];then $ECHO;fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

PARALLEL_SUMMARY_STATUS_REPORT () {
D
Daniel Gustafsson 已提交
1051
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
1052 1053 1054 1055 1056 1057 1058 1059 1060 1061 1062 1063 1064 1065 1066 1067 1068 1069 1070 1071 1072 1073 1074 1075
	REPORT_FAIL=0
	if [ -f $1 ];then
	        KILLED_COUNT=`$GREP -c "KILLED:" $PARALLEL_STATUS_FILE`
                COMPLETED_COUNT=`$GREP -c "COMPLETED:" $PARALLEL_STATUS_FILE`
                FAILED_COUNT=`$GREP -c "FAILED:" $PARALLEL_STATUS_FILE`
		((TOTAL_FAILED_COUNT=$KILLED_COUNT+$FAILED_COUNT))
                LOG_MSG "[INFO]:------------------------------------------------" 1
                LOG_MSG "[INFO]:-Parallel process exit status" 1
                LOG_MSG "[INFO]:------------------------------------------------" 1
                LOG_MSG "[INFO]:-Total processes marked as completed           = $COMPLETED_COUNT" 1
                if [ $KILLED_COUNT -ne 0 ];then
                LOG_MSG "[WARN]:-Total processes marked as killed              = $KILLED_COUNT $WARN_MARK" 1
		REPORT_FAIL=1
                else
                LOG_MSG "[INFO]:-Total processes marked as killed              = 0" 1
                fi
                if [ $FAILED_COUNT -ne 0 ];then
                LOG_MSG "[WARN]:-Total processes marked as failed              = $FAILED_COUNT $WARN_MARK" 1
		REPORT_FAIL=1
                else
                LOG_MSG "[INFO]:-Total processes marked as failed              = 0" 1
                fi
                LOG_MSG "[INFO]:------------------------------------------------" 1
	else
S
Shoaib Lari 已提交
1076
		LOG_MSG "[WARN]:-Could not locate status file $1" 1
1077 1078 1079 1080 1081 1082
		REPORT_FAIL=1
	fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

CHK_GPDB_ID () {
D
Daniel Gustafsson 已提交
1083
	LOG_MSG "[INFO]:-Start Function $FUNCNAME"
1084 1085 1086 1087 1088 1089 1090 1091 1092 1093 1094 1095 1096 1097 1098 1099 1100 1101 1102 1103 1104 1105
	if [ -f ${INITDB} ];then
	        PERMISSION=`ls -al ${INITDB}|$AWK '{print $1}'`
		MASTER_INITDB_ID=`ls -al ${INITDB}|$AWK '{print $3}'`
		INIT_CHAR=`$ECHO $MASTER_INITDB_ID|$TR -d '\n'|$WC -c|$TR -d ' '`
		MASTER_INITDB_GROUPID=`ls -al ${INITDB}|$AWK '{print $4}'`
		GROUP_INIT_CHAR=`$ECHO $MASTER_INITDB_ID|$TR -d '\n'|$WC -c|$TR -d ' '`
		GPDB_ID=`id|$TR '(' ' '|$TR ')' ' '|$AWK '{print $2}'`
		GPDB_GROUPID=`id|$TR '(' ' '|$TR ')' ' '|$AWK '{print $4}'`

		USER_EXECUTE=`$ECHO $PERMISSION | $SED -e 's/...\(.\).*/\1/g'`
		GROUP_EXECUTE=`$ECHO $PERMISSION | $SED -e 's/......\(.\).*/\1/g'`

		if [ `$ECHO $GPDB_ID|$TR -d '\n'|$WC -c` -gt $INIT_CHAR ];then
			GPDB_ID_CHK=`$ECHO $GPDB_ID|$CUT -c1-$INIT_CHAR`
		else
			GPDB_ID_CHK=$GPDB_ID
		fi

		if [ `$ECHO $GPDB_GROUPID|$TR -d '\n'|$WC -c` -gt $GROUP_INIT_CHAR ];then
			GPDB_GROUPID_CHK=`$ECHO $GPDB_GROUPID|$CUT -c1-$GROUP_INIT_CHAR`
		else
			GPDB_GROUPID_CHK=$GPDB_GROUPID
D
Daniel Gustafsson 已提交
1106
		fi
1107

D
Daniel Gustafsson 已提交
1108
		if [ x$GPDB_ID_CHK == x$MASTER_INITDB_ID ] && [ x"x" == x"$USER_EXECUTE" ];then
1109 1110 1111 1112 1113 1114 1115 1116 1117 1118 1119 1120 1121 1122 1123 1124 1125
		    LOG_MSG "[INFO]:-Current user id of $GPDB_ID, matches initdb id of $MASTER_INITDB_ID"
		elif [ x$GPDB_GROUPID_CHK == x$MASTER_INITDB_GROUPID ] && [ x"x" == x"$GROUP_EXECUTE" ] ; then
		    LOG_MSG "[INFO]:-Current group id of $GPDB_GROUPID, matches initdb group id of $MASTER_INITDB_GROUPID"
		else
			LOG_MSG "[WARN]:-File permission mismatch.  The $GPDB_ID_CHK owns the Greenplum Database installation directory."
			LOG_MSG "[WARN]:-You are currently logged in as $MASTER_INITDB_ID and may not have sufficient"
			LOG_MSG "[WARN]:-permissions to run the Greenplum binaries and management utilities."
		fi

		if [ x"" != x"$USER" ];then
			if [ `$ECHO $USER|$TR -d '\n'|$WC -c` -gt $INIT_CHAR ];then
				USER_CHK=`$ECHO $USER|$CUT -c1-$INIT_CHAR`
			else
				USER_CHK=$USER
			fi
			if [ x$GPDB_ID_CHK != x$USER_CHK ];then
				LOG_MSG "[WARN]:-\$USER mismatch, id returns $GPDB_ID, \$USER returns $USER" 1
D
Daniel Gustafsson 已提交
1126
				LOG_MSG "[WARN]:-The GPDB superuser account that owns the initdb binary should run these utilities" 1
1127 1128 1129 1130 1131 1132 1133 1134 1135 1136 1137 1138 1139 1140
				LOG_MSG "[WARN]:-This may cause problems when these utilities are run as $USER" 1
			fi
		else
			LOG_MSG "[INFO]:-Environment variable \$USER unset, will set to $GPDB_ID" 1
			export USER=$GPDB_ID
		fi
		if [ x"" != x"$LOGNAME" ];then
			if [ `$ECHO $LOGNAME|$TR -d '\n'|$WC -c` -gt $INIT_CHAR ];then
				LOGNAME_CHK=`$ECHO $LOGNAME|$CUT -c1-$INIT_CHAR`
			else
				LOGNAME_CHK=$LOGNAME
			fi
			if [ x$GPDB_ID_CHK != x$LOGNAME_CHK ];then
				LOG_MSG "[WARN]:-\$LOGNAME mismatch, id returns $GPDB_ID_CHK, \$LOGNAME returns $LOGNAME_CHK" 1
D
Daniel Gustafsson 已提交
1141
				LOG_MSG "[WARN]:-The GPDB superuser account that owns the initdb binary should run these utilities" 1
1142 1143 1144 1145 1146 1147 1148 1149 1150 1151 1152 1153 1154 1155 1156 1157 1158 1159 1160 1161 1162 1163 1164 1165 1166 1167 1168 1169
				LOG_MSG "[WARN]:-This may cause problems when these utilities are run as $LOGNAME" 1
			fi
		else
			LOG_MSG "[INFO]:-Environment variable \$LOGNAME unset, will set to $GPDB_ID" 1
			export LOGNAME=$GPDB_ID
		fi
	else
		LOG_MSG "[WARN]:-No initdb file, unable to verify id" 1
	fi
	LOG_MSG "[INFO]:-End Function $FUNCNAME"
}

#******************************************************************************
# Main Section
#******************************************************************************
#******************************************************************************
# Setup logging directory
#******************************************************************************
CUR_DATE=`$DATE +%Y%m%d`
DEFLOGDIR=$HOME/gpAdminLogs
if [ ! -d $DEFLOGDIR ]; then
		mkdir $DEFLOGDIR
fi
LOG_FILE=$DEFLOGDIR/${PROG_NAME}_${CUR_DATE}.log

#Set up OS type for scripts to change command lines
OS_TYPE=`uname -s|tr '[A-Z]' '[a-z]'`
case $OS_TYPE in
1170
	linux ) IPV4_ADDR_LIST_CMD="`findCmdInPath ip` -4 address show"
1171 1172 1173 1174 1175 1176 1177 1178 1179
		IPV6_ADDR_LIST_CMD="`findCmdInPath ip` -6 address show"
		PS_TXT="ax"
		LIB_TYPE="LD_LIBRARY_PATH"
		PG_METHOD="ident"
		HOST_ARCH_TYPE="uname -i"
		NOLINE_ECHO="$ECHO -e"
		DEFAULT_LOCALE_SETTING=en_US.utf8
		PING6=`findCmdInPath ping6`
		PING_TIME="-c 1"
1180
		;;
1181
	darwin ) IPV4_ADDR_LIST_CMD="$IFCONFIG -a inet"
1182 1183 1184 1185 1186 1187 1188 1189
		IPV6_ADDR_LIST_CMD="$IFCONFIG -a inet6"
		PS_TXT="ax"
		LIB_TYPE="DYLD_LIBRARY_PATH"
		# Darwin zcat wants to append ".Z" to the end of the file name; use "gunzip -c" instead
		PG_METHOD="ident"
		HOST_ARCH_TYPE="uname -m"
		NOLINE_ECHO=$ECHO
		DEFAULT_LOCALE_SETTING=en_US.utf-8
1190
		PING6=`findCmdInPath ping6`
1191
		PING_TIME="-c 1"
1192
		;;
1193 1194
	freebsd ) IPV4_ADDR_LIST_CMD="$IFCONFIG -a inet"
		IPV6_ADDR_LIST_CMD="$IFCONFIG -a inet6"
1195 1196 1197 1198 1199 1200
		LIB_TYPE="LD_LIBRARY_PATH"
		PG_METHOD="ident"
		HOST_ARCH_TYPE="uname -m"
		NOLINE_ECHO="$ECHO -e"
		DEFAULT_LOCALE_SETTING=en_US.utf8
		PING_TIME="-c 1"
1201
		;;
1202 1203 1204 1205 1206 1207 1208 1209 1210
	openbsd ) IPV4_ADDR_LIST_CMD="ifconfig -a inet"
		IPV6_ADDR_LIST_CMD="ifconfig -a inet6"
		LIB_TYPE="LD_LIBRARY_PATH"
		PG_METHOD="ident"
		HOST_ARCH_TYPE="uname -m"
		NOLINE_ECHO="echo -e"
		DEFAULT_LOCALE_SETTING=en_US.UTF-8
		PING_TIME="-c 1"
		DF="df -P"
1211
		;;
1212 1213 1214 1215 1216 1217 1218 1219 1220 1221 1222 1223 1224 1225
	* ) echo unknown ;;
esac

GP_LIBRARY_PATH=`$DIRNAME \`$DIRNAME $INITDB\``/lib

##
# we setup some EXPORT foo='blah' commands for when we dispatch to segments and standby master
##
EXPORT_GPHOME='export GPHOME='$GPHOME
if [ x"$LIB_TYPE" == x"LD_LIBRARY_PATH" ]; then
    EXPORT_LIB_PATH="export LD_LIBRARY_PATH=$LD_LIBRARY_PATH"
else
    EXPORT_LIB_PATH="export DYLD_LIBRARY_PATH=$DYLD_LIBRARY_PATH"
fi