pacemaker-cli-utils 1.1.16-1 / usr / sbin / crm_report

#!/bin/sh

 # Copyright (C) 2010 Andrew Beekhof <andrew@beekhof.net>
 #
 # This program is free software; you can redistribute it and/or
 # modify it under the terms of the GNU General Public
 # License as published by the Free Software Foundation; either
 # version 2.1 of the License, or (at your option) any later version.
 #
 # This software is distributed in the hope that it will be useful,
 # but WITHOUT ANY WARRANTY; without even the implied warranty of
 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
 # General Public License for more details.
 #
 # You should have received a copy of the GNU General Public
 # License along with this library; if not, write to the Free Software
 # Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
 #


# Note the quotes around `$TEMP': they are essential!
TEMP=`getopt				\
    -o hv?xl:f:t:n:T:L:p:c:dSACHu:D:MVse:	\
    --long help,cts:,cts-log:,dest:,node:,nodes:,from:,to:,sos-mode,logfile:,as-directory,single-node,cluster:,user:,max-depth:,version,features,rsh:	\
    -n 'crm_report' -- "$@"`
eval set -- "$TEMP"

progname=$(basename "$0")
rsh="ssh -T"
times=""
tests=""
nodes=""
compress=1
cluster="any"
ssh_user="root"
search_logs=1
report_data=`dirname $0`
maxdepth=5

extra_logs=""
sanitize_patterns="passw.*"
log_patterns="CRIT: ERROR:"

usage() {
cat<<EOF
$progname - Create a tarball containing everything needed when reporting cluster problems


Usage: $progname --from date-time [--to date-time] [additional options] [dest]

Options:
  -V			increase verbosity
  -v,--version          software version
     --features         software features
  -f, --from time       time prior to problems beginning: "YYYY-M-D H:M:S"  (do not forget the quotes)
  -t, --to time	        time at which all problems were resolved: "YYYY-M-D H:M:S" (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 the cluster is not active on the current machine
			accepts both -n "a b" and -n a -n b
  -l, --logfile file	log file to collect, normally this will be determined automatically
  -p patt		additional regular expression to match variables to be removed
			(default: "passw.*")
  -L patt		additional regular expression to match in log files for analysis
			(default: $log_patterns)
  -M			collect only the logs specified by -l
  -S, --single-node	single node operation; don't try to start report collectors on other nodes
  -c, --cluster type	force the cluster type (corosync,openais,heartbeat,logmaster)
  -A, --openais		force the cluster type to be OpenAIS
  -C, --corosync	force the cluster type to be CoroSync
  -H, --heartbeat	force the cluster type to be Heartbeat
  -u, --user user	ssh username for cluster nodes (default: root)
  -D, --depth           search depth to use when attempting to locate files
  -e, --rsh		specify the remote shell to use (default: ssh -T)
  --sos-mode            use defaults suitable for being called by sosreport tool
                        (behavior subject to change and not useful to end users)
  --dest       		a custom destination directory/file
  dest			a custom destination directory/file

$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 ssh.

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 1.1.16-94ff4df"; exit 0;;
    --features)     echo "1.1.16-94ff4df:  generated-manpages agent-manpages ncurses libqb-logging libqb-ipc lha-fencing systemd nagios  corosync-native atomic-attrd snmp libesmtp acls"; 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=/usr/share/pacemaker
else
    echo "Using local helpers"
fi

. $report_data/report.common

while true; do
    case "$1" in
	-x) set -x; shift;;
	-V) verbose=`expr $verbose + 1`; shift;;
	-T|--cts-test) tests="$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;;
	-E|-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;;
	-A|--openais)   cluster="openais";   shift;;
	-C|--corosync)  cluster="corosync";  shift;;
	-H|--heartbeat) cluster="heartbeat"; shift;;
	-c|--cluster)   cluster="$2"; shift; shift;;
	-e|--rsh)       rsh="$2";     shift; shift;;
	-u|--user)      ssh_user="$2"; shift; shift;;
        -D|--max-depth)     maxdepth="$2"; shift; shift;;
	-M) search_logs=0; shift;;
        --sos-mode) search_logs=0; nodes="$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=$search_logs
verbose=$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 "    http://bugs.clusterlabs.org/enter_bug.cgi?product=Pacemaker"
	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 -a -f $d2/RUNNING ] ||
	[ -f $d1/STOPPED -a -f $d2/STOPPED ]; 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;;
	$B_CONF) diff -u $1 $2;; # confdiff?
	*)       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="$HOSTCACHE $MEMBERSHIP_F $CIB_F $CRM_MON_F $B_CONF logd.cf $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 "/var/lib/pacemaker/cib/cib.xml" ]; then
	debug "Querying on-disk CIB for nodes"
        grep "node " "/var/lib/pacemaker/cib/cib.xml" | node_names_from_xml
        return
    fi

    # 3. hostcache
    if [ -z "$HA_STATE_DIR" ]; then
        HA_STATE_DIR=/var/lib/heartbeat
    fi
    if [ -f "$HA_STATE_DIR/hostcache" ]; then
	debug "Reading nodes from $HA_STATE_DIR/hostcache"
        awk '{print $1}' "$HA_STATE_DIR/hostcache"
        return
    fi

    # 4. ha.cf
    if [ "x$cluster" = "xheartbeat" ]; then
        cluster_cf=$(find_cluster_cf $cluster)
	debug "Reading nodes from $cluster_cf"
        getcfvar $cluster node "$cluster_cf"
        return
    fi

    # 5. logs
    # TODO: This has multiple issues:
    # * It looks for messages from crm_update_peer(), which is used only by
    #   heartbeat and legacy plugin clusters; it should work with CMAN and
    #   corosync2 clusters as well.
    # * It does a findmsg for "crm_update_peer" (which will hit
    #   "crm_update_peer_proc" etc.), but then greps for "crm_update_peer:".
    # * It always uses grep, even though $logfile might be compressed.
    #   For this reason and efficiency, it would nice if findmsg could
    #   optionally print the matches instead of the file names.
    # * It would be nice to skip this step for Pacemaker Remote nodes since their
    #   logs will not have node names, but it is nontrivial to know that.
    #   Cluster nodes generally won't get here, but stopped Pacemaker Remote
    #   nodes will.
    logfile=$(findmsg 1 "crm_update_peer")
    debug "Looking for nodes in $logfile"
    if [ ! -z "$logfile" ]; then
        grep crm_update_peer: "$logfile" \
            | sed s/.*crm_update_peer// \
            | sed s/://g \
            | awk '{print $2}' \
            | grep -v "(null)" \
            | sort -u \
            | tr '\n' ' '
    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 "crmd\\|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: