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); then
  61     if [[ $lock_info != *non-blanked* ]]; then
  62       locked=true
  63     fi
  64   else
  65     locked=true
  66   fi
  67 }
  68
  69 usage() {
  70   cat <<EOF
  71 Usage: ${0##*/} [OPTIONS]
  72 Do btrfs maintence or stop if we have X and xprintidle shows a user
  73
  74 Normally, no options are needed.
  75
  76 --check  Only check if an existing maintence should be cancelled due to
  77          nonidle user and run in a loop every 20 seconds for 10
  78          minutes.
  79
  80 --dryrun Just print out what we would do.
  81
  82 --force  Run regardless of user idle status on all disks and do scrub
  83          regardless of when it was last run.
  84 --no-stats  Avoid checking error statistics. Use this to avoid a rare race
  85             condition when running --check concurrently with normal run.
  86
  87
  88 -h|--help   Show help
  89
  90 Note: Uses util-linux getopt option parsing: spaces between args and
  91 options, short options can be combined, options before args.
  92 EOF
  93   exit $1
  94 }
  95
  96 ##### begin command line parsing ########
  97
  98 # ensure we can handle args with spaces or empty.
  99 ret=0; getopt -T || ret=$?
 100 [[ $ret == 4 ]] || { echo "Install util-linux for enhanced getopt" >&2; exit 1; }
 101
 102 check=false
 103 dryrun=false
 104 force=false
 105 stats=true
 106
 107 temp=$(getopt -l help,check,dryrun,force,no-stats h "$@") || usage 1
 108 eval set -- "$temp"
 109 while true; do
 110   case $1 in
 111     --check) check=true ;;
 112     --dryrun) dryrun=true ;;
 113     --force) force=true ;;
 114     --no-stats) stats=false ;;
 115     -h|--help) usage ;;
 116     --) shift; break ;;
 117     *) echo "$0: unexpected args: $*" >&2 ; usage 1 ;;
 118   esac
 119   shift
 120 done
 121 readonly check dryrun force stats
 122 ##### end command line parsing ########
 123
 124
 125 main() {
 126   if ! $force; then
 127     check-idle
 128     if ! $check; then
 129       min=0
 130       max_min=300
 131       # When the cron kicks in, we may not be idle (physically sleeping) yet, so
 132       # wait.
 133       while ! $locked && (( min < max_min )); do
 134         min=$(( min + 1 ))
 135         sleep 60
 136         check-idle
 137       done
 138       # If we've waited a really long time for idle, just give up.
 139       if (( min == max_min )); then
 140         return
 141       fi
 142     fi
 143   fi
 144
 145
 146   fnd="findmnt --types btrfs --noheading"
 147   for x in $($fnd --output "SOURCE" --nofsroot | sort -u); do
 148     mnt=$($fnd --output "TARGET" --first-only --source $x)
 149     [[ $mnt ]] || continue
 150
 151     #### begin look for diff in stats, eg: increasing error count ####
 152     if $stats; then
 153       tmp=$(mktemp)
 154       # ${mnt%/} so that if mnt is / we avoid making a buggy looking path
 155       stats_path=${mnt%/}/btrfs-dev-stats
 156       if [[ ! -e $stats_path ]]; then
 157         btrfs dev stats -c $mnt >$stats_path ||: # populate initial reading
 158       elif ! btrfs dev stats -c $mnt >$tmp; then
 159         if ! diff -q $stats_path $tmp; then
 160           mv $stats_path $stats_path.1
 161           cat $tmp >$stats_path
 162           diff=$(diff -u $stats_path $tmp 2>&1 ||:)
 163           printf "diff of: btrfs dev stats -c %s\n%s\n" "$mnt" "$diff"
 164           exim -odf -i root <<EOF
 165 From: root@$(hostname -f)
 166 To: root@$(hostname -f)
 167 Subject: btrfsmaint: device stats changed for $mnt
 168
 169 diff of: btrfs dev stats -c $mnt
 170 $diff
 171 EOF
 172         fi
 173       fi
 174       rm -f $tmp
 175     fi
 176     #### end look for diff in stats, eg: increasing error count ####
 177
 178     if $check; then
 179       if ! $locked; then
 180         if $dryrun; then
 181           echo "$0: not idle. if this wasnt a dry run, btrfs scrub cancel $mnt"
 182         else
 183           btrfs scrub cancel $mnt &>/dev/null ||:
 184         fi
 185       fi
 186       continue
 187     fi
 188
 189     # for comparing before and after balance.
 190     # the log is already fairly verbose, so commented.
 191     # e btrfs filesystem df $mnt
 192     # e df -H $mnt
 193     if btrfs filesystem df $mnt | grep -q "Data+Metadata"; then
 194       for usage in $dusage; do
 195         e ionice -c 3 btrfs balance start -dusage=$usage -musage=$usage $mnt
 196       done
 197     else
 198       e ionice -c 3 btrfs balance start -dusage=0 $mnt
 199       for usage in $dusage; do
 200         e ionice -c 3 btrfs balance start -dusage=$usage $mnt
 201       done
 202       e ionice -c 3 btrfs balance start -musage=0 $mnt
 203       for usage in $musage; do
 204         e ionice -c 3 btrfs balance start -musage=$usage $mnt
 205       done
 206     fi
 207     date=
 208     scrub_status=$(btrfs scrub status $mnt)
 209     if printf "%s\n" "$scrub_status" | grep -i '^status:[[:space:]]*finished$' &>/dev/null; then
 210       date=$(printf "%s\n" "$scrub_status" | sed -rn 's/^Scrub started:[[:space:]]*(.*)/\1/p')
 211     fi
 212     if [[ ! $date ]]; then
 213       # output from older versions, at least btrfs v4.15.1
 214       date=$(
 215         printf "%s\n" "$scrub_status" | \
 216           sed -rn 's/^\s*scrub started at (.*) and finished.*/\1/p'
 217           )
 218     fi
 219     if ! $force && [[ $date ]]; then
 220       if $dryrun; then
 221         echo "$0: last scrub finish for $mnt: $date"
 222       fi
 223       date=$(date --date="$date" +%s)
 224       # if date is sooner than 60 days ago
 225       # the wiki recommends 30 days or so, but
 226       # I'm going with 60 days.
 227       if (( date > EPOCHSECONDS - 60*60*24*60 )); then
 228         if $dryrun; then
 229           echo "$0: skiping scrub of $mnt, last was $(( (EPOCHSECONDS - date) / 60/60/24 )) days ago, < 30 days"
 230         fi
 231         continue
 232       fi
 233     fi
 234     # btrfsmaintenance does -c 2 -n 4, but I want lowest pri.
 235     e btrfs scrub start -Bd -c 3 $mnt
 236
 237     # We normally only do one disk since this is meant to be run in
 238     # downtime and if we try to do all disks, we invariably end up doing
 239     # a scrub after downtime. So, just do one disk per day.
 240     if ! $force; then
 241       return 0
 242     fi
 243   done
 244 }
 245
 246 loop-main() {
 247   while true; do
 248     main
 249     sleep 60
 250   done
 251 }
 252
 253 if $check; then
 254   loop-main
 255 else
 256   main
 257 fi