#!/bin/sh

#

# Copyright 2010-2019 the Pacemaker project contributors

#

# The version control history for this file may have further details.

#

# This source code is licensed under the GNU General Public License version 2

# or later (GPLv2+) WITHOUT ANY WARRANTY.

#

TEMP=`@GETOPT_PATH@			\

    -o hv?xl:f:t:n:T:L:p:c:dSCu:D:MVse:	\

    --long help,corosync,cts:,cts-log:,dest:,node:,nodes:,from:,to:,sos-mode,logfile:,as-directory,single-node,cluster:,user:,max-depth:,version,features,rsh:	\

    -n 'crm_report' -- "$@"`

# The quotes around $TEMP are essential

eval set -- "$TEMP"

progname=$(basename "$0")

rsh="ssh -T"

tests=""

nodes=""

compress=1

cluster="any"

ssh_user="root"

search_logs=1

sos_mode=0

report_data=`dirname $0`

maxdepth=5

extra_logs=""

sanitize_patterns="passw.*"

log_patterns="CRIT: ERROR:"

usage() {

cat<

$progname - Create archive of everything needed when reporting cluster problems

Usage: $progname [options] [DEST]

Required option:

  -f, --from TIME       time prior to problems beginning

                        (as "YYYY-M-D H:M:S" including the quotes)

Options:

  -V                    increase verbosity (may be specified multiple times)

  -h, --help            display this message

  -v, --version         display software version

  --features            display software features

  -t, --to TIME         time at which all problems were resolved

                        (as "YYYY-M-D H:M:S" including the quotes; default "now")

  -T, --cts TEST        CTS test or set of tests to extract

  --cts-log             CTS master logfile

  -n, --nodes NODES     node names for this cluster (only needed if cluster is

                        not active on this host; accepts -n "a b" or -n a -n b)

  -M                    do not search for cluster logs

  -l, --logfile FILE    log file to collect (in addition to detected logs if -M

                        is not specified; may be specified multiple times)

  -p PATT               additional regular expression to match variables to be

                        masked in output (default: "passw.*")

  -L PATT               additional regular expression to match in log files for

                        analysis (default: $log_patterns)

  -S, --single-node     don't attempt to collect data from other nodes

  -c, --cluster TYPE    force the cluster type instead of detecting

                        (currently only corosync is supported)

  -C, --corosync        force the cluster type to be corosync

  -u, --user USER       username to use when collecting data from other nodes

                        (default root)

  -D, --max-depth       search depth to use when attempting to locate files

  -e, --rsh             command to use to run commands on other nodes

                        (default ssh -T)

  -d, --as-directory    leave result as a directory tree instead of archiving

  --sos-mode            use defaults suitable for being called by sosreport tool

                        (behavior subject to change and not useful to end users)

  DEST, --dest DEST     custom destination directory or file name

$progname works best when run from a cluster node on a running cluster,

but can be run from a stopped cluster node or a Pacemaker Remote node.

If neither --nodes nor --single-node is given, $progname will guess the

node list, but may have trouble detecting Pacemaker Remote nodes.

Unless --single-node is given, the node names (whether specified by --nodes

or detected automatically) must be resolvable and reachable via the command

specified by -e/--rsh using the user specified by -u/--user.

Examples:

   $progname -f "2011-12-14 13:05:00" unexplained-apache-failure

   $progname -f 2011-12-14 -t 2011-12-15 something-that-took-multiple-days

   $progname -f 13:05:00   -t 13:12:00   brief-outage

EOF

}

case "$1" in

    -v|--version)   echo "$progname @VERSION@-@BUILD_VERSION@"; exit 0;;

    --features)     echo "@VERSION@-@BUILD_VERSION@: @PCMK_FEATURES@"; exit 0;;

    --|-h|--help) usage; exit 0;;

esac

# Prefer helpers in the same directory if they exist, to simplify development

if [ ! -f $report_data/report.common ]; then

    report_data=@datadir@/@PACKAGE@

else

    echo "Using local helpers"

fi

. $report_data/report.common

while true; do

    case "$1" in

	-x) set -x; shift;;

	-V) verbose=`expr $verbose + 1`; shift;;

	-T|--cts) tests="$tests $2"; shift; shift;;

	   --cts-log) ctslog="$2"; shift; shift;;

	-f|--from) start_time=`get_time "$2"`; shift; shift;;

	-t|--to) end_time=`get_time "$2"`; shift; shift;;

	-n|--node|--nodes) nodes="$nodes $2"; shift; shift;;

	-S|--single-node) nodes="$host"; shift;;

	-l|--logfile) extra_logs="$extra_logs $2"; shift; shift;;

	-p) sanitize_patterns="$sanitize_patterns $2"; shift; shift;;

	-L) log_patterns="$log_patterns `echo $2 | sed 's/ /\\\W/g'`"; shift; shift;;

	-d|--as-directory) compress=0; shift;;

	-C|--corosync)  cluster="corosync";  shift;;

	-c|--cluster)   cluster="$2"; shift; shift;;

	-e|--rsh)       rsh="$2";     shift; shift;;

	-u|--user)      ssh_user="$2"; shift; shift;;

        -D|--max-depth)     maxdepth="$2"; shift; shift;;

	-M) search_logs=0; shift;;

        --sos-mode) sos_mode=1; nodes="$host"; shift;;

	--dest) DESTDIR=$2; shift; shift;;

	--) if [ ! -z $2 ]; then DESTDIR=$2; fi; break;;

	-h|--help) usage; exit 0;;

	# Options for compatibility with hb_report

	-s) shift;;

	*) echo "Unknown argument: $1"; usage; exit 1;;

    esac

done

collect_data() {

    label="$1"

    start=`expr $2 - 10`

    end=`expr $3 + 10`

    masterlog=$4

    if [ "x$DESTDIR" != x ]; then

	echo $DESTDIR | grep -e "^/" -qs

	if [ $? = 0 ]; then

	    l_base=$DESTDIR

	else

	    l_base="`pwd`/$DESTDIR"

	fi

	debug "Using custom scratch dir: $l_base"

	r_base=`basename $l_base`

    else

	l_base=$HOME/$label

	r_base=$label

    fi

    if [ -e $l_base ]; then

	fatal "Output directory $l_base already exists, specify an alternate name with --dest"

    fi

    mkdir -p $l_base

    if [ "x$masterlog" != "x" ]; then

	dumplogset "$masterlog" $start $end > "$l_base/$HALOG_F"

    fi

    for node in $nodes; do

	cat <<EOF >$l_base/.env

LABEL="$label"

REPORT_HOME="$r_base"

REPORT_MASTER="$host"

REPORT_TARGET="$node"

LOG_START=$start

LOG_END=$end

REMOVE=1

SANITIZE="$sanitize_patterns"

CLUSTER=$cluster

LOG_PATTERNS="$log_patterns"

EXTRA_LOGS="$extra_logs"

SEARCH_LOGS=$search_logs

SOS_MODE=$sos_mode

verbose=$verbose

maxdepth=$maxdepth

EOF

	if [ $host = $node ]; then

	    cat <<EOF >>$l_base/.env

REPORT_HOME="$l_base"

EOF

	    cat $l_base/.env $report_data/report.common $report_data/report.collector > $l_base/collector

	    bash $l_base/collector

	else

	    cat $l_base/.env $report_data/report.common $report_data/report.collector \

		| $rsh -l $ssh_user $node -- "mkdir -p $r_base; cat > $r_base/collector; bash $r_base/collector" | (cd $l_base && tar mxf -)

	fi

    done

    analyze $l_base > $l_base/$ANALYSIS_F

    if [ -f $l_base/$HALOG_F ]; then

	node_events $l_base/$HALOG_F > $l_base/$EVENTS_F

    fi

    for node in $nodes; do

	cat $l_base/$node/$ANALYSIS_F >> $l_base/$ANALYSIS_F

	if [ -s $l_base/$node/$EVENTS_F ]; then

	    cat $l_base/$node/$EVENTS_F >> $l_base/$EVENTS_F

	elif [ -s $l_base/$HALOG_F ]; then

	    awk "\$4==\"$nodes\"" $l_base/$EVENTS_F >> $l_base/$n/$EVENTS_F

	fi

    done

    log " "

    if [ $compress = 1 ]; then

	fname=`shrink $l_base`

	rm -rf $l_base

	log "Collected results are available in $fname"

	log " "

	log "Please create a bug entry at"

	log "    @BUG_URL@"

	log "Include a description of your problem and attach this tarball"

	log " "

	log "Thank you for taking time to create this report."

    else

	log "Collected results are available in $l_base"

    fi

    log " "

}

#

# check if files have same content in the cluster

#

cibdiff() {

    d1=$(dirname $1)

    d2=$(dirname $2)

    if [ -f "$d1/RUNNING" ] && [ ! -f "$d2/RUNNING" ]; then

        DIFF_OK=0

    elif [ -f "$d1/STOPPED" ] && [ ! -f "$d2/STOPPED" ]; then

        DIFF_OK=0

    else

        DIFF_OK=1

    fi

    if [ $DIFF_OK -eq 1 ]; then

	if which crm_diff > /dev/null 2>&1; then

	    crm_diff -c -n $1 -o $2

	else

	    info "crm_diff(8) not found, cannot diff CIBs"

	fi

    else

	echo "can't compare cibs from running and stopped systems"

    fi

}

diffcheck() {

    [ -f "$1" ] || {

	echo "$1 does not exist"

	return 1

    }

    [ -f "$2" ] || {

	echo "$2 does not exist"

	return 1

    }

    case $(basename "$1") in

        $CIB_F)  cibdiff $1 $2 ;;

        *)       diff -u $1 $2 ;;

    esac

}

#

# remove duplicates if files are same, make links instead

#

consolidate() {

    for n in $nodes; do

	if [ -f $1/$2 ]; then

	    rm $1/$n/$2

	else

	    mv $1/$n/$2 $1

	fi

	ln -s ../$2 $1/$n

    done

}

analyze_one() {

    rc=0

    node0=""

    for n in $nodes; do

	if [ "$node0" ]; then

	    diffcheck $1/$node0/$2 $1/$n/$2

	    rc=$(($rc+$?))

	else

	    node0=$n

	fi

    done

    return $rc

}

analyze() {

    flist="$MEMBERSHIP_F $CIB_F $CRM_MON_F $SYSINFO_F"

    for f in $flist; do

	printf "Diff $f... "

	ls $1/*/$f >/dev/null 2>&1 || {

	    echo "no $1/*/$f :/"

	    continue

	}

	if analyze_one $1 $f; then

	    echo "OK"

	    [ "$f" != $CIB_F ] && consolidate $1 $f

	else

	    echo ""

	fi

    done

}

do_cts() {

    test_sets=`echo $tests | tr ',' ' '`

    for test_set in $test_sets; do

	start_time=0

	start_test=`echo $test_set | tr '-' ' ' | awk '{print $1}'`

	end_time=0

	end_test=`echo $test_set | tr '-' ' ' | awk '{print $2}'`

	if [ x$end_test = x ]; then

	    msg="Extracting test $start_test"

	    label="CTS-$start_test-`date +"%b-%d-%Y"`"

	    end_test=`expr $start_test + 1`

	else

	    msg="Extracting tests $start_test to $end_test"

	    label="CTS-$start_test-$end_test-`date +"%b-%d-%Y"`"

	    end_test=`expr $end_test + 1`

	fi

	if [ $start_test = 0 ]; then

	    start_pat="BEGINNING [0-9].* TESTS"

	else

	    start_pat="Running test.*\[ *$start_test\]"

	fi

	if [ x$ctslog = x ]; then

	    ctslog=`findmsg 1 "$start_pat"`

	    if [ x$ctslog = x ]; then

		fatal "No CTS control file detected"

	    else

		log "Using CTS control file: $ctslog"

	    fi

	fi

	line=`grep -n "$start_pat" $ctslog | tail -1 | sed 's/:.*//'`

	if [ ! -z "$line" ]; then

	    start_time=`linetime $ctslog $line`

	fi

	line=`grep -n "Running test.*\[ *$end_test\]" $ctslog | tail -1 | sed 's/:.*//'`

	if [ ! -z "$line" ]; then

	    end_time=`linetime $ctslog $line`

	fi

	if [ -z "$nodes" ]; then

	    nodes=`grep CTS: $ctslog | grep -v debug: | grep " \* " | sed s:.*\\\*::g | sort -u  | tr '\\n' ' '`

	    log "Calculated node list: $nodes"

	fi

	if [ $end_time -lt $start_time ]; then

	    debug "Test didn't complete, grabbing everything up to now"

	    end_time=`date +%s`

	fi

	if [ $start_time != 0 ];then

	    log "$msg (`time2str $start_time` to `time2str $end_time`)"

	    collect_data $label $start_time $end_time $ctslog

	else

	    fatal "$msg failed: not found"

	fi

    done

}

node_names_from_xml() {

    awk '

      /uname/ {

            for( i=1; i<=NF; i++ )

                    if( $i~/^uname=/ ) {

                            sub("uname=.","",$i);

                            sub("\".*","",$i);

                            print $i;

                            next;

                    }

      }

    ' | tr '\n' ' '

}

getnodes() {

    cluster="$1"

    # 1. Live (cluster nodes or Pacemaker Remote nodes)

    # TODO: This will not detect Pacemaker Remote nodes unless they

    # have ever had a permanent node attribute set, because it only

    # searches the nodes section. It should also search the config

    # for resources that create Pacemaker Remote nodes.

    cib_nodes=$(cibadmin -Ql -o nodes 2>/dev/null)

    if [ $? -eq 0 ]; then

	debug "Querying CIB for nodes"

        echo "$cib_nodes" | node_names_from_xml

        return

    fi

    # 2. Saved

    if [ -f "@CRM_CONFIG_DIR@/cib.xml" ]; then

	debug "Querying on-disk CIB for nodes"

        grep "node " "@CRM_CONFIG_DIR@/cib.xml" | node_names_from_xml

        return

    fi

    # 3. logs

    # TODO: Look for something like crm_update_peer

}

if [ $compress -eq 1 ]; then

    require_tar

fi

if [ "x$tests" != "x" ]; then

    do_cts

elif [ "x$start_time" != "x" ]; then

    masterlog=""

    if [ -z "$sanitize_patterns" ]; then

	log "WARNING: The tarball produced by this program may contain"

	log "         sensitive information such as passwords."

	log ""

	log "We will attempt to remove such information if you use the"

	log "-p option. For example: -p \"pass.*\" -p \"user.*\""

	log ""

	log "However, doing this may reduce the ability for the recipients"

	log "to diagnose issues and generally provide assistance."

	log ""

	log "IT IS YOUR RESPONSIBILITY TO PROTECT SENSITIVE DATA FROM EXPOSURE"

	log ""

    fi

    # If user didn't specify a cluster stack, make a best guess if possible.

    if [ -z "$cluster" ] || [ "$cluster" = "any" ]; then

        cluster=$(get_cluster_type)

    fi

    # If user didn't specify node(s), make a best guess if possible.

    if [ -z "$nodes" ]; then

	nodes=`getnodes $cluster`

        if [ -n "$nodes" ]; then

            log "Calculated node list: $nodes"

        else

            fatal "Cannot determine nodes; specify --nodes or --single-node"

        fi

    fi

    if

	echo $nodes | grep -qs $host

    then

	debug "We are a cluster node"

    else

	debug "We are a log master"

	masterlog=`findmsg 1 "pacemaker-controld\\|CTS"`

    fi

    if [ -z $end_time ]; then

	end_time=`perl -e 'print time()'`

    fi

    label="pcmk-`date +"%a-%d-%b-%Y"`"

    log "Collecting data from $nodes (`time2str $start_time` to `time2str $end_time`)"

    collect_data $label $start_time $end_time $masterlog

else

    fatal "Not sure what to do, no tests or time ranges to extract"

fi

# vim: set expandtab tabstop=8 softtabstop=4 shiftwidth=4 textwidth=80: