misc/mdcheck

   1 #!/bin/bash
   2
   3 # Copyright (C) 2014 Neil Brown <neilb@suse.de>
   4 #
   5 #
   6 #    This program is free software; you can redistribute it and/or modify
   7 #    it under the terms of the GNU General Public License as published by
   8 #    the Free Software Foundation; either version 2 of the License, or
   9 #    (at your option) any later version.
  10 #
  11 #    This program is distributed in the hope that it will be useful,
  12 #    but WITHOUT ANY WARRANTY; without even the implied warranty of
  13 #    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  14 #    GNU General Public License for more details.
  15 #
  16 #    Author: Neil Brown
  17 #    Email: <neilb@suse.de>
  18
  19 # This script should be run periodically to automatically
  20 # perform a 'check' on any md arrays.
  21 #
  22 # It supports a 'time budget' such that any incomplete 'check'
  23 # will be checkpointed when that time has expired.
  24 # A subsequent invocation can allow the 'check' to continue.
  25 #
  26 # Options are:
  27 #   --continue    Don't start new checks, only continue old ones.
  28 #   --duration    This is passed to "date --date=$duration" to find out
  29 #                 when to finish
  30 #
  31 # To support '--continue', arrays are identified by UUID and the 'sync_completed'
  32 # value is stored  in /var/lib/mdcheck/$UUID
  33
  34 # convert a /dev/md name into /sys/.../md equivalent
  35 sysname() {
  36         set `ls -lLd $1`
  37         maj=${5%,}
  38         min=$6
  39         readlink -f /sys/dev/block/$maj:$min
  40 }
  41
  42 args=$(getopt -o hcd: -l help,continue,duration: -n mdcheck -- "$@")
  43 rv=$?
  44 if [ $rv -ne 0 ]; then exit $rv; fi
  45
  46 eval set -- $args
  47
  48 cont=
  49 endtime=
  50 while [ " $1" != " --" ]
  51 do
  52     case $1 in
  53         --help )
  54                 echo >&2 'Usage: mdcheck [--continue] [--duration time-offset]'
  55                 echo >&2 '  time-offset must be understood by "date --date"'
  56                 exit 0
  57                 ;;
  58         --continue ) cont=yes ;;
  59         --duration ) shift; dur=$1
  60                 endtime=$(date --date "$dur" "+%s")
  61                 ;;
  62     esac
  63     shift
  64 done
  65 shift
  66
  67 # We need a temp file occasionally...
  68 tmp=/var/lib/mdcheck/.md-check-$$
  69 trap 'rm -f "$tmp"' 0
  70
  71
  72 # firstly, clean out really old state files
  73 mkdir -p /var/lib/mdcheck
  74 find /var/lib/mdcheck -name "MD_UUID*" -type f -mtime +180 -exec rm {} \;
  75
  76 # Now look at each md device.
  77 cnt=0
  78 for dev in /dev/md?*
  79 do
  80         [ -e "$dev" ] || continue
  81         sys=`sysname $dev`
  82         if [ ! -f "$sys/md/sync_action" ]
  83         then # cannot check this array
  84                 continue
  85         fi
  86         if [ "`cat $sys/md/sync_action`" != 'idle' ]
  87         then # This array is busy
  88                 continue
  89         fi
  90
  91         mdadm --detail --export "$dev" | grep '^MD_UUID=' > $tmp || continue
  92         source $tmp
  93         fl="/var/lib/mdcheck/MD_UUID_$MD_UUID"
  94         if [ -z "$cont" ]
  95         then
  96                 start=0
  97         elif [ -z "$MD_UUID" -o ! -f "$fl" ]
  98         then
  99                 # Nothing to continue here
 100                 continue
 101         else
 102                 start=`cat "$fl"`
 103         fi
 104
 105         cnt=$[cnt+1]
 106         eval MD_${cnt}_fl=\$fl
 107         eval MD_${cnt}_sys=\$sys
 108         echo $start > $fl
 109         echo $start > $sys/md/sync_min
 110         echo check > $sys/md/sync_action
 111 done
 112
 113 if [ -z "$endtime" ]
 114 then
 115         exit 0
 116 fi
 117
 118 while [ `date +%s` -lt $endtime ]
 119 do
 120         any=
 121         for i in `eval echo {1..$cnt}`
 122         do
 123                 eval fl=\$MD_${i}_fl
 124                 eval sys=\$MD_${i}_sys
 125
 126                 if [ -z "$fl" ]; then continue; fi
 127
 128                 if [ "`cat $sys/md/sync_action`" != 'check' ]
 129                 then
 130                         eval MD_${i}_fl=
 131                         rm -f $fl
 132                         continue;
 133                 fi
 134                 read a rest < $sys/md/sync_completed
 135                 echo $a > $fl
 136                 any=yes
 137         done
 138         if [ -z "$any" ]; then exit 0; fi
 139         sleep 120
 140 done
 141
 142 # We've waited, and there are still checks running.
 143 # Time to stop them.
 144 for i in `eval echo {1..$cnt}`
 145 do
 146         eval fl=\$MD_${i}_fl
 147         eval sys=\$MD_${i}_sys
 148
 149         if [ -z "$fl" ]; then continue; fi
 150
 151         if [ "`cat $sys/md/sync_action`" != 'check' ]
 152         then
 153                 eval MD_${i}_fl=
 154                 rm -f $fl
 155                 continue;
 156         fi
 157         echo idle > $sys/md/sync_action
 158         cat $sys/md/sync_min > $fl
 159 done