contrib-munin/plugins/emc/emc_vnx_block_lun_perfdata

#!/bin/bash

: <<=cut

=head1 NAME 

 emc_vnx_block_lun_perfdata - Plugin to monitor Block statistics of EMC VNX 5300
 Unified Storage Processors

=head1 AUTHOR

 Evgeny Beysembaev <megabotva@gmail.com>

=head1 LICENSE

 GPLv2

=head1 MAGIC MARKERS

  #%# family=auto
  #%# capabilities=autoconf

=head1 DESCRIPTION

 The plugin monitors LUN of EMC Unified Storage FLARE SP's. Probably it can also
 be compatible with  other Clariion systems. It uses SSH to connect to Control 
 Stations, then remotely executes /nas/sbin/navicli and fetches and parses data 
 from it. Obviously, it's easy to reconfigure plugin not to use Control Stations'
 navicli in favor of using locally installed /opt/Navisphere's cli. There is no 
 difference which Storage Processor to use to gather data, so this plugin tries 
 both of them and uses the first active one. This plugin also automatically 
 chooses Primary Control Station from the list by calling /nasmcd/sbin/getreason 
 and /nasmcd/sbin/t2slot.
 
 I left some parts of this plugin as rudimental to make easy to reconfigure it 
 to draw more (or less) data.

 The plugin has been tested in the following Operating Environment (OE):
  File Version T7.1.76.4
  Block Revision 05.32.000.5.215

=head1 COMPATIBILITY

 The plugin has been written for being compatible with EMC VNX5300 Storage 
 system, as this is the only EMC storage which i have. By the way, i am pretty 
 sure it can also work with other VNX1 storages, like VNX5100 and VNX5500, and 
 old-style Clariion systems.
 About VNX2 series, i don't know whether the plugin will be able to work with 
 them. Maybe it would need some corrections in command-line backend. The same 
 situation is with other EMC systems, so i encourage you to try and fix the 
 plugin. 
 
=head1 LIST OF GRAPHS

 Graph category Disk:
	EMC VNX 5300 LUN Blocks
	EMC VNX 5300 LUN Requests
	EMC VNX 5300 Counted Load per LUN
	EMC VNX 5300 Sum of Outstanding Requests
	EMC VNX 5300 Non-Zero Request Count Arrivals
	EMC VNX 5300 Trespasses
	EMC VNX 5300 Counted Block Queue Length
	EMC VNX 5300 Counted Load per SP


=head1 CONFIGURATION

=head2 Prerequisites

 First of all, be sure that statistics collection is turned on. You can do this
 by typing:
 navicli -h spa setstats -on
 on your Control Station or locally through /opt/Navisphere 

 Also, the plugin actively uses buggy "cdef" feature of Munin 2.0, and here we 
 can be hit by the following bugs:
 http://munin-monitoring.org/ticket/1017 - Here I have some workarounds in the 
 plugin, be sure that they are working.
 http://munin-monitoring.org/ticket/1352 - Metrics in my plugin can be much 
 longer than 15 characters.
 Without these workarounds "Load" and "Queue Length" would not work.

=head2 Installation

 The plugin uses SSH to connect to Control Stations. It's possible to use 
 'nasadmin' user, but it would be better if you create read-only global user by
 Unisphere Client. The user should have only Operator role. 
 I created "operator" user but due to the fact that Control Stations already 
 had one internal "operator" user, the new one was called "operator1". So be 
 careful. After that, copy .bash_profile from /home/nasadmin to a newly created
 /home/operator1.
 
 On munin-node side choose a user which will be used to connect through SSH. 
 Generally user "munin" is ok. Then, execute "sudo su munin -s /bin/bash", 
 "ssh-keygen" and "ssh-copy-id" to both Control Stations with newly created 
 user.
 
 Make a link from /usr/share/munin/plugins/emc_vnx_dm_basic_stats to 
 /etc/munin/plugins/emc_vnx_dm_basic_stats_<NAME>, where <NAME> is any 
 arbitrary name of your storage system. The plugin will return <NAME> in its 
 answer as "host_name" field.
 
 For example, assume your storage system is called "VNX5300".
 Make a configuration file at 
 /etc/munin/plugin-conf.d/emc_vnx_block_lun_perfdata_VNX5300. For example:
 
 [emc_vnx_block_lun_perfdata_VNX5300]
 user munin					
 env.username operator1
 env.cs_addr 192.168.1.1 192.168.1.2	

 or:

 [emc_vnx_block_lun_perfdata_VNX5300]
 user munin                                     
 env.username operator1
 env.localcli /opt/Navisphere/bin/naviseccli
 env.sp_addr 192.168.0.3 192.168.0.4
 env.blockpw foobar

 Where: 
 user - SSH Client local user
 env.username - Remote user with Operator role for Block or File part
 env.cs_addr - Control Stations addresses for remote (indirect) access.
 env.localcli - Optional. Path of localhost 'Naviseccli' binary. If this 
   variable is set, env.cs_addr is ignored, and local 'navicli' is used.
   Requires env.blockpw variable.
 env.sp_addr - Default is "SPA SPB". In case of "direct" connection to 
   Storage Processors, their addresses/hostnames are written here.
 env.blockpw - Password for connecting to Storage Processors

=head1 ERRATA

 It counts Queue Length in not fully correct way. We take parameters totally 
 from both SP's, but after we divide them independently by load of SPA and SPB.
 Anyway, in most AAA / ALUA cases the formula is correct.

=head1 HISTORY

 09.11.2016 - First Release
 26.12.2016 - Compatibility with Munin coding style

=cut

export LANG=C

. "$MUNIN_LIBDIR/plugins/plugin.sh"

cs_addr="${cs_addr:=""}"
username="${username:=""}"
blockpw="${blockpw:=""}"

TARGET=$(echo "${0##*/}" | cut -d _ -f 6)

# "All Storage Processors we have"
if [[ -v "sp_addr" ]]; then
	SPALL=$sp_addr
else
	SPALL="SPA SPB"
fi
# "navicli" command. Can be local or remote, through Control Stations
if [[ -v "localcli" ]]; then
	NAVICLI=$localcli
else
	NAVICLI="/nas/sbin/navicli"
fi

# Prints "10" on stdout if found Primary Online control station. "11" - for Secondary Online control station. 
ssh_check_cmd() {
		ssh -q "$username@$1" "/nasmcd/sbin/getreason | grep -w \"slot_\$(/nasmcd/sbin/t2slot)\" | cut -d- -f1 | awk '{print \$1}' "
}


check_conf_and_set_vars () {
	if [ -z "$username" ]; then
		echo "No username ('username' environment variable)!"
		return 1
	fi

	if [ -z "$localcli" ]; then
		if [ -z "$cs_addr" ]; then
			echo "No control station addresses ('cs_addr' environment variable)!"
			return 1
		fi
		#Choosing Cotrol Station. Code have to be "10"
		for CS in $cs_addr; do
			# shellcheck disable=SC2086
			if [[ "10" = "$(ssh_check_cmd $CS)" ]]; then
				PRIMARY_CS=$CS
				SSH="ssh -q $username@$PRIMARY_CS "
				break
			fi
		done

		if [ -z "$PRIMARY_CS" ]; then
			echo "No alive primary Control Station from list \"$cs_addr\"";
			return 1
		fi
	else
		if [ ! -f "$localcli" ]; then
			echo "Local CLI is set, but no binary found at $localcli!"
			return 1
		fi
		if [ -z "$blockpw" ]; then
			echo "No Password for Block Access ('blockpw' environment variable)!"
			return 1
		fi
		SSH=""
		NAVICLI="$localcli -User $username -Password $blockpw -Scope 0 "
	fi
	local probe_sp
	for probe_sp in $SPALL; do
		# shellcheck disable=SC2086
		if $SSH $NAVICLI -h "$probe_sp" >/dev/null 2>&1; then
			StorageProcessor="$probe_sp"
			break
		fi
	done
	[ -z "$StorageProcessor" ] && echo "No active Storage Processor found!" && return 1
	NAVICLI_NOSP="$NAVICLI -h"
	NAVICLI="$NAVICLI -h $StorageProcessor"
	return 0
}

if [ "$1" = "autoconf" ]; then
	check_conf_ans=$(check_conf_and_set_vars)
		if [ $? -eq 0 ]; then
			echo "yes"
		else
			echo "no ($check_conf_ans)"
		fi
		exit 0
fi

check_conf_and_set_vars 1>&2 || exit 1

run_remote() {
	if [ -z "$SSH" ]; then
		sh -c "$*"
	else
		$SSH "$*"
	fi
}

run_navicli() {
	run_remote "$NAVICLI" "$*"
}


# Get Lun List
LUNLIST=$(run_navicli lun -list -drivetype | sed -ne 's/^Name:\ *//p' | sort)

echo "host_name ${TARGET}"
echo

if [ "$1" = "config" ] ; then
	cat <<-EOF 
		multigraph emc_vnx_block_blocks
		graph_category disk
		graph_title EMC VNX 5300 LUN Blocks
		graph_vlabel Blocks Read (-) / Written (+)
		graph_args --base 1000
	EOF

	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		cat <<-EOF 
			${LUN}_read.label none
			${LUN}_read.graph no
			${LUN}_read.min 0
			${LUN}_read.draw AREA
			${LUN}_read.type COUNTER
			${LUN}_write.label $LUN Blocks
			${LUN}_write.negative ${LUN}_read
			${LUN}_write.type COUNTER
			${LUN}_write.min 0
			${LUN}_write.draw STACK
		EOF
	done <<< "$LUNLIST"

	cat <<-EOF

		multigraph emc_vnx_block_req
		graph_category disk
		graph_title EMC VNX 5300 LUN Requests
		graph_vlabel Requests: Read (-) / Write (+)
		graph_args --base 1000
	EOF
	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		cat <<-EOF
			${LUN}_readreq.label none
			${LUN}_readreq.graph no
			${LUN}_readreq.min 0
			${LUN}_readreq.type COUNTER
			${LUN}_writereq.label $LUN Requests
			${LUN}_writereq.negative ${LUN}_readreq
			${LUN}_writereq.type COUNTER
			${LUN}_writereq.min 0
		EOF
	done <<< "$LUNLIST"

	cat <<-EOF

		multigraph emc_vnx_block_ticks
		graph_category disk
		graph_title EMC VNX 5300 Counted Load per LUN
		graph_vlabel Load, % * Number of LUNs 
		graph_args --base 1000 -l 0 -r 
	EOF
	echo -n "graph_order "
	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		echo -n "${LUN}_busyticks ${LUN}_idleticks ${LUN}_bta=${LUN}_busyticks_spa ${LUN}_idleticks_spa ${LUN}_btb=${LUN}_busyticks_spb ${LUN}_idleticks_spb "
	done <<< "$LUNLIST"
	echo ""
	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		cat <<-EOF
			${LUN}_busyticks_spa.label $LUN Busy Ticks SPA
			${LUN}_busyticks_spa.type COUNTER
			${LUN}_busyticks_spa.graph no
			${LUN}_bta.label $LUN Busy Ticks SPA
			${LUN}_bta.graph no
			${LUN}_idleticks_spa.label $LUN Idle Ticks SPA
			${LUN}_idleticks_spa.type COUNTER
			${LUN}_idleticks_spa.graph no
			${LUN}_busyticks_spb.label $LUN Busy Ticks SPB
			${LUN}_busyticks_spb.type COUNTER
			${LUN}_busyticks_spb.graph no
			${LUN}_btb.label $LUN Busy Ticks SPB
			${LUN}_btb.graph no
			${LUN}_idleticks_spb.label $LUN Idle Ticks SPB
			${LUN}_idleticks_spb.type COUNTER
			${LUN}_idleticks_spb.graph no
			${LUN}_load_spa.label $LUN load SPA 
			${LUN}_load_spa.draw AREASTACK
			${LUN}_load_spb.label $LUN load SPB
			${LUN}_load_spb.draw AREASTACK
			${LUN}_load_spa.cdef 100,${LUN}_bta,${LUN}_busyticks_spa,${LUN}_idleticks_spa,+,/,*
			${LUN}_load_spb.cdef 100,${LUN}_btb,${LUN}_busyticks_spa,${LUN}_idleticks_spa,+,/,*
		EOF
	done <<< "$LUNLIST"

	cat <<-EOF
		
		multigraph emc_vnx_block_outstanding
		graph_category disk
		graph_title EMC VNX 5300 Sum of Outstanding Requests
		graph_vlabel Requests
		graph_args --base 1000
	EOF
	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		cat <<-EOF 
			${LUN}_outstandsum.label $LUN
			${LUN}_outstandsum.type COUNTER
		EOF
	done <<< "$LUNLIST"

	cat <<-EOF
		
		multigraph emc_vnx_block_nonzeroreq
		graph_category disk
		graph_title EMC VNX 5300 Non-Zero Request Count Arrivals
		graph_vlabel Count Arrivals
		graph_args --base 1000
	EOF
	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		cat <<-EOF
			${LUN}_nonzeroreq.label $LUN
			${LUN}_nonzeroreq.type COUNTER
		EOF
	done <<< "$LUNLIST"

	cat <<-EOF

		multigraph emc_vnx_block_trespasses
		graph_category disk
		graph_title EMC VNX 5300 Trespasses
		graph_vlabel Trespasses
	EOF
	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		cat <<-EOF
			${LUN}_implic_tr.label ${LUN} Implicit Trespasses
			${LUN}_explic_tr.label ${LUN} Explicit Trespasses
		EOF
	done <<< "$LUNLIST"

	cat <<-EOF

		multigraph emc_vnx_block_queue
		graph_category disk
		graph_title EMC VNX 5300 Counted Block Queue Length 
		graph_vlabel Length
	EOF
	while read -r LUN ; do
		LUN="$(clean_fieldname "$LUN")"
		cat <<-EOF
			${LUN}_busyticks_spa.label ${LUN}
			${LUN}_busyticks_spa.graph no
			${LUN}_busyticks_spa.type COUNTER
			${LUN}_idleticks_spa.label ${LUN}
			${LUN}_idleticks_spa.graph no
			${LUN}_idleticks_spa.type COUNTER
			${LUN}_busyticks_spb.label ${LUN}
			${LUN}_busyticks_spb.graph no
			${LUN}_busyticks_spb.type COUNTER
			${LUN}_idleticks_spb.label ${LUN}
			${LUN}_idleticks_spb.graph no
			${LUN}_idleticks_spb.type COUNTER
			${LUN}_outstandsum.label ${LUN}
			${LUN}_outstandsum.graph no
			${LUN}_outstandsum.type COUNTER
			${LUN}_nonzeroreq.label ${LUN}
			${LUN}_nonzeroreq.graph no
			${LUN}_nonzeroreq.type COUNTER
			${LUN}_readreq.label ${LUN}
			${LUN}_readreq.graph no
			${LUN}_readreq.type COUNTER
			${LUN}_writereq.label ${LUN}
			${LUN}_writereq.graph no
			${LUN}_writereq.type COUNTER
		EOF
		# Queue Length SPA = ((Sum of Outstanding Requests SPA - NonZero Request Count Arrivals SPA / 2)/(Host Read Requests SPA + Host Write Requests SPA))*
		# (Busy Ticks SPA/(Busy Ticks SPA + Idle Ticks SPA)
		# We count together SPA and SPB, although it is not fully corrext
		cat <<-EOF
			${LUN}_ql_l_a.label ${LUN} Queue Length SPA
			${LUN}_ql_l_a.cdef ${LUN}_outstandsum,${LUN}_nonzeroreq,2,/,-,${LUN}_readreq,${LUN}_writereq,+,/,${LUN}_busyticks_spa,*,${LUN}_busyticks_spa,${LUN}_idleticks_spa,+,/
			${LUN}_ql_l_b.label ${LUN} Queue Length SPB
			${LUN}_ql_l_b.cdef ${LUN}_outstandsum,${LUN}_nonzeroreq,2,/,-,${LUN}_readreq,${LUN}_writereq,+,/,${LUN}_busyticks_spb,*,${LUN}_busyticks_spb,${LUN}_idleticks_spb,+,/
		EOF
	done <<< "$LUNLIST"
	cat <<-EOF

		multigraph emc_vnx_block_ticks_total
		graph_category disk
		graph_title EMC VNX 5300 Counted Load per SP
		graph_vlabel Load, %
	EOF
	echo -n "graph_order "
	for SP in $SPALL; do
		SPclean="$(clean_fieldname "$SP")"
		echo -n "${SPclean}_total_bt=${SPclean}_total_busyticks "
	done
	echo ""
	for SP in $SPALL; do
		SPclean="$(clean_fieldname "$SP")"
		cat <<-EOF
			${SPclean}_total_busyticks.label ${SP}
			${SPclean}_total_busyticks.graph no
			${SPclean}_total_busyticks.type COUNTER 
			${SPclean}_total_bt.label ${SP}
			${SPclean}_total_bt.graph no
			${SPclean}_total_bt.type COUNTER 
			${SPclean}_total_idleticks.label ${SP}
			${SPclean}_total_idleticks.graph no
			${SPclean}_total_idleticks.type COUNTER
			${SPclean}_total_load.label ${SP} Total Load
			${SPclean}_total_load.cdef ${SPclean}_total_bt,${SPclean}_total_busyticks,${SPclean}_total_idleticks,+,/,100,*
		EOF
	done
	exit 0
fi

#Preparing big complex command to SP's to have most work done remotely.
#BIGCMD="$SSH"
while read -r LUN ; do
	FILTERLUN="$(clean_fieldname "$LUN")"
	BIGCMD+="$NAVICLI lun -list -name $LUN -perfData | 
		sed -ne 's/^Blocks Read\:\ */${FILTERLUN}_read.value /p; 
		s/^Blocks Written\:\ */${FILTERLUN}_write.value /p;
		s/Read Requests\:\ */${FILTERLUN}_readreq.value /p;
		s/Write Requests\:\ */${FILTERLUN}_writereq.value /p;
		s/Busy Ticks SP A\:\ */${FILTERLUN}_busyticks_spa.value /p;
		s/Idle Ticks SP A\:\ */${FILTERLUN}_idleticks_spa.value /p;
		s/Busy Ticks SP B\:\ */${FILTERLUN}_busyticks_spb.value /p;
		s/Idle Ticks SP B\:\ */${FILTERLUN}_idleticks_spb.value /p;
		s/Sum of Outstanding Requests\:\ */${FILTERLUN}_outstandsum.value /p;
		s/Non-Zero Request Count Arrivals\:\ */${FILTERLUN}_nonzeroreq.value /p;
		s/Implicit Trespasses\:\ */${FILTERLUN}_implic_tr.value /p;
		s/Explicit Trespasses\:\ */${FILTERLUN}_explic_tr.value /p;
		' ;  "
done <<< "$LUNLIST"
ANSWER=$(run_remote "$BIGCMD")

for SP in $SPALL; do
	FILTER_SP="$(clean_fieldname "$SP")"
	BIGCMD="getcontrol -cbt | sed -ne '
		s/Controller busy ticks\:\ */${FILTER_SP}_total_busyticks.value /p;
		s/Controller idle ticks\:\ */${FILTER_SP}_total_idleticks.value /p;
		'
	"
	ANSWER+=$'\n'$(run_remote "$NAVICLI_NOSP $SP" "$BIGCMD")
done

get_precise_answer_field() {
    echo "$ANSWER" | grep -F "_${1}."
}

echo "multigraph emc_vnx_block_blocks"
get_precise_answer_field "read"
get_precise_answer_field "write"
echo -e "\nmultigraph emc_vnx_block_req"
get_precise_answer_field "readreq"
get_precise_answer_field "writereq"

echo -e "\nmultigraph emc_vnx_block_ticks"
while read -r LUN ; do
	LUN="$(clean_fieldname "$LUN")"
	#Will count these values later, using cdef
	echo "${LUN}_load_spa.value 0"
	echo "${LUN}_load_spb.value 0"
done <<< "$LUNLIST"
get_precise_answer_field "busyticks_spa"
get_precise_answer_field "idleticks_spa"
get_precise_answer_field "busyticks_spb"
get_precise_answer_field "idleticks_spb"

echo -e "\nmultigraph emc_vnx_block_outstanding"
get_precise_answer_field "outstandsum"

echo -e "\nmultigraph emc_vnx_block_nonzeroreq"
get_precise_answer_field "nonzeroreq"

echo -e "\nmultigraph emc_vnx_block_trespasses"
get_precise_answer_field "implic_tr"
get_precise_answer_field "explic_tr"

echo -e "\nmultigraph emc_vnx_block_queue"
# Queue Length
get_precise_answer_field "busyticks_spa"
get_precise_answer_field "idleticks_spa"
get_precise_answer_field "busyticks_spb"
get_precise_answer_field "idleticks_spb"
get_precise_answer_field "outstandsum"
get_precise_answer_field "nonzeroreq"
get_precise_answer_field "readreq"
get_precise_answer_field "writereq"
while read -r LUN ; do
	LUN="$(clean_fieldname "$LUN")"
	#Will count these values later, using cdef
	echo "${LUN}_ql_l_a.value 0 "
	echo "${LUN}_ql_l_b.value 0 "
done <<< "$LUNLIST"

echo -e "\nmultigraph emc_vnx_block_ticks_total"
get_precise_answer_field "total_busyticks"
get_precise_answer_field "total_idleticks"
#Will count them later
for SP in $SPALL; do
	SP="$(clean_fieldname "$SP")"
	echo "${SP}_total_load.value 0"
done

exit 0