iankelling.org Git - distro-setup/blob - btrfsmaint

   1 #!/bin/bash
   2 # I, Ian Kelling, follow the GNU license recommendations at
   3 # https://www.gnu.org/licenses/license-recommendations.en.html. They
   4 # recommend that small programs, < 300 lines, be licensed under the
   5 # Apache License 2.0. This file contains or is part of one or more small
   6 # programs. If a small program grows beyond 300 lines, I plan to switch
   7 # its license to GPL.
   8
   9 # Copyright 2024 Ian Kelling
  10
  11 # Licensed under the Apache License, Version 2.0 (the "License");
  12 # you may not use this file except in compliance with the License.
  13 # You may obtain a copy of the License at
  14
  15 #     http://www.apache.org/licenses/LICENSE-2.0
  16
  17 # Unless required by applicable law or agreed to in writing, software
  18 # distributed under the License is distributed on an "AS IS" BASIS,
  19 # WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
  20 # See the License for the specific language governing permissions and
  21 # limitations under the License.
  22
  23
  24
  25 [[ $EUID == 0 ]] || exec sudo -E "${BASH_SOURCE[0]}" "$@"
  26
  27 set -e; . /usr/local/lib/bash-bear; set +e
  28
  29 # inspired from
  30 # https://github.com/kdave/btrfsmaintenance
  31
  32 if [[ $INVOCATION_ID ]]; then
  33   err-cleanup() {
  34     exim -odf -i root <<EOF
  35 From: root@$(hostname -f)
  36 To: root@$(hostname -f)
  37 Subject: btrfsmaint automatically exited on command error
  38
  39 journalctl -u btrfsmaint -n 50:
  40 $(journalctl -u btrfsmaint -n 50)
  41 EOF
  42   }
  43 fi
  44
  45 dusage="5 10"
  46 musage="5"
  47
  48 e() {
  49   echo "btrfsmaint: $*"
  50   if ! $dryrun; then
  51     "$@"
  52   fi
  53 }
  54
  55 check-idle() {
  56   type -p xscreensaver-command &>/dev/null || return 0
  57   export XAUTHORITY=/home/iank/.Xauthority
  58   export DISPLAY=:0
  59   locked=false
  60   if lock_info=$(xscreensaver-command -time 2>/dev/null); then
  61     if [[ $lock_info != *non-blanked* ]]; then
  62       locked=true
  63     fi
  64   fi
  65 }
  66
  67 usage() {
  68   cat <<EOF
  69 Usage: ${0##*/} [OPTIONS]
  70 Do btrfs maintence or stop if we have X and xprintidle shows a user
  71
  72 Normally, no options are needed.
  73
  74 --check  Only check if an existing maintence should be cancelled due to
  75          nonidle user and run in a loop every 20 seconds for 10
  76          minutes.
  77
  78 --dryrun Just print out what we would do.
  79
  80 --force  Run regardless of user idle status on all disks and do scrub
  81          regardless of when it was last run.
  82 --no-stats  Avoid checking error statistics. Use this to avoid a rare race
  83             condition when running --check concurrently with normal run.
  84
  85
  86 -h|--help   Show help
  87
  88 Note: Uses util-linux getopt option parsing: spaces between args and
  89 options, short options can be combined, options before args.
  90 EOF
  91   exit $1
  92 }
  93
  94 ##### begin command line parsing ########
  95
  96 # ensure we can handle args with spaces or empty.
  97 ret=0; getopt -T || ret=$?
  98 [[ $ret == 4 ]] || { echo "Install util-linux for enhanced getopt" >&2; exit 1; }
  99
 100 check=false
 101 dryrun=false
 102 force=false
 103 stats=true
 104
 105 temp=$(getopt -l help,check,dryrun,force,no-stats h "$@") || usage 1
 106 eval set -- "$temp"
 107 while true; do
 108   case $1 in
 109     --check) check=true ;;
 110     --dryrun) dryrun=true ;;
 111     --force) force=true ;;
 112     --no-stats) stats=false ;;
 113     -h|--help) usage ;;
 114     --) shift; break ;;
 115     *) echo "$0: unexpected args: $*" >&2 ; usage 1 ;;
 116   esac
 117   shift
 118 done
 119 readonly check dryrun force stats
 120 ##### end command line parsing ########
 121
 122
 123 main() {
 124   if ! $force; then
 125     check-idle
 126     if ! $check; then
 127       min=0
 128       max_min=300
 129       # When the cron kicks in, we may not be idle (physically sleeping) yet, so
 130       # wait.
 131       while ! $locked && (( min < max_min )); do
 132         min=$(( min + 1 ))
 133         sleep 60
 134         check-idle
 135       done
 136       # If we've waited a really long time for idle, just give up.
 137       if (( min == max_min )); then
 138         return
 139       fi
 140     fi
 141   fi
 142
 143
 144   fnd="findmnt --types btrfs --noheading"
 145   for x in $($fnd --output "SOURCE" --nofsroot | sort -u); do
 146     mnt=$($fnd --output "TARGET" --first-only --source $x)
 147     [[ $mnt ]] || continue
 148
 149     #### begin look for diff in stats, eg: increasing error count ####
 150     if $stats; then
 151       tmp=$(mktemp)
 152       # ${mnt%/} so that if mnt is / we avoid making a buggy looking path
 153       stats_path=${mnt%/}/btrfs-dev-stats
 154       if [[ ! -e $stats_path ]]; then
 155         btrfs dev stats -c $mnt >$stats_path ||: # populate initial reading
 156       elif ! btrfs dev stats -c $mnt >$tmp; then
 157         if ! diff -q $stats_path $tmp; then
 158           mv $stats_path $stats_path.1
 159           cat $tmp >$stats_path
 160           diff=$(diff -u $stats_path $tmp 2>&1 ||:)
 161           printf "diff of: btrfs dev stats -c %s\n%s\n" "$mnt" "$diff"
 162           exim -odf -i root <<EOF
 163 From: root@$(hostname -f)
 164 To: root@$(hostname -f)
 165 Subject: btrfsmaint: device stats changed for $mnt
 166
 167 diff of: btrfs dev stats -c $mnt
 168 $diff
 169 EOF
 170         fi
 171       fi
 172       rm -f $tmp
 173     fi
 174     #### end look for diff in stats, eg: increasing error count ####
 175
 176     if $check; then
 177       if ! $locked; then
 178         if $dryrun; then
 179           echo "$0: not idle. if this wasnt a dry run, btrfs scrub cancel $mnt"
 180         else
 181           btrfs scrub cancel $mnt &>/dev/null ||:
 182         fi
 183       fi
 184       continue
 185     fi
 186
 187     # for comparing before and after balance.
 188     # the log is already fairly verbose, so commented.
 189     # e btrfs filesystem df $mnt
 190     # e df -H $mnt
 191     if btrfs filesystem df $mnt | grep -q "Data+Metadata"; then
 192       for usage in $dusage; do
 193         e ionice -c 3 btrfs balance start -dusage=$usage -musage=$usage $mnt
 194       done
 195     else
 196       e ionice -c 3 btrfs balance start -dusage=0 $mnt
 197       for usage in $dusage; do
 198         e ionice -c 3 btrfs balance start -dusage=$usage $mnt
 199       done
 200       e ionice -c 3 btrfs balance start -musage=0 $mnt
 201       for usage in $musage; do
 202         e ionice -c 3 btrfs balance start -musage=$usage $mnt
 203       done
 204     fi
 205     date=
 206     scrub_status=$(btrfs scrub status $mnt)
 207     if printf "%s\n" "$scrub_status" | grep -i '^status:[[:space:]]*finished$' &>/dev/null; then
 208       date=$(printf "%s\n" "$scrub_status" | sed -rn 's/^Scrub started:[[:space:]]*(.*)/\1/p')
 209     fi
 210     if [[ ! $date ]]; then
 211       # output from older versions, at least btrfs v4.15.1
 212       date=$(
 213         printf "%s\n" "$scrub_status" | \
 214           sed -rn 's/^\s*scrub started at (.*) and finished.*/\1/p'
 215           )
 216     fi
 217     if ! $force && [[ $date ]]; then
 218       if $dryrun; then
 219         echo "$0: last scrub finish for $mnt: $date"
 220       fi
 221       date=$(date --date="$date" +%s)
 222       # if date is sooner than 60 days ago
 223       # the wiki recommends 30 days or so, but
 224       # I'm going with 60 days.
 225       if (( date > EPOCHSECONDS - 60*60*24*60 )); then
 226         if $dryrun; then
 227           echo "$0: skiping scrub of $mnt, last was $(( (EPOCHSECONDS - date) / 60/60/24 )) days ago, < 30 days"
 228         fi
 229         continue
 230       fi
 231     fi
 232     # btrfsmaintenance does -c 2 -n 4, but I want lowest pri.
 233     e btrfs scrub start -Bd -c 3 $mnt
 234
 235     # We normally only do one disk since this is meant to be run in
 236     # downtime and if we try to do all disks, we invariably end up doing
 237     # a scrub after downtime. So, just do one disk per day.
 238     if ! $force; then
 239       return 0
 240     fi
 241   done
 242 }
 243
 244 loop-main() {
 245   while true; do
 246     main
 247     sleep 60
 248   done
 249 }
 250
 251 if $check; then
 252   loop-main
 253 else
 254   main
 255 fi