X-Git-Url: https://iankelling.org/git/?a=blobdiff_plain;f=btrfsmaint;h=0204ed5aa78a4e9043e0aae3a624066252a77955;hb=72c18f3a6a7f1ed0ca16af654a1f804ab96e1ff9;hp=8e7214187a13e7709880f6d6d1fd169fed22074e;hpb=da085f71e5ae00a9f78ed903be6ae675f66eacfa;p=distro-setup diff --git a/btrfsmaint b/btrfsmaint index 8e72141..0204ed5 100755 --- a/btrfsmaint +++ b/btrfsmaint @@ -3,20 +3,29 @@ [[ $EUID == 0 ]] || exec sudo -E "${BASH_SOURCE[0]}" "$@" -source /a/bin/errhandle/err +set -e; . /usr/local/lib/err; set +e # inspired from # https://github.com/kdave/btrfsmaintenance +if [[ $INVOCATION_ID ]]; then + err-cleanup() { + exim -odf -i root <&2; exit 1; } -force=false check=false dryrun=false -if [[ $1 ]]; then +force=false +stats=true + +temp=$(getopt -l help,check,dryrun,force,no-stats h "$@") || usage 1 +eval set -- "$temp" +while true; do case $1 in - check) - check=true - ;; - force) - force=true - ;; - dryrun) - dryrun=true - ;; - *) - echo "$0: error: unexpected arg" >&2 - usage 1 - ;; + --check) check=true ;; + --dryrun) dryrun=true ;; + --force) force=true ;; + --no-stats) stats=false ;; + -h|--help) usage ;; + --) shift; break ;; + *) echo "$0: unexpected args: $*" >&2 ; usage 1 ;; esac -fi + shift +done +readonly check dryrun force stats +##### end command line parsing ######## main() { idle=true if ! $force; then check-idle + if ! $check; then + min=0 + max_min=300 + # When the cron kicks in, we may not be idle (physically sleeping) yet, so + # wait. + while ! $idle && (( min < max_min )); do + min=$(( min + 1 )) + sleep 60 + check-idle + done + # If we've waited a really long time for idle, just give up. + if (( min == max_min )); then + return + fi + fi fi - tmp=$(mktemp) fnd="findmnt --types btrfs --noheading" for x in $($fnd --output "SOURCE" --nofsroot | sort -u); do mnt=$($fnd --output "TARGET" --first-only --source $x) [[ $mnt ]] || continue - if ! btrfs dev stats -c $mnt >$tmp; then - if diff -q $mnt/btrfs-dev-stats $tmp; then - diff -u $mnt/btrfs-dev-stats $tmp | mail -s "$HOSTNAME: error: btrfs dev stats -c $mnt" root@localhost - cat $tmp >$mnt/btrfs-dev-stats + #### begin look for diff in stats, eg: increasing error count #### + if $stats; then + tmp=$(mktemp) + # ${mnt%/} so that if mnt is / we avoid making a buggy looking path + stats_path=${mnt%/}/btrfs-dev-stats + if [[ ! -e $stats_path ]]; then + btrfs dev stats -c $mnt >$stats_path ||: # populate initial reading + elif ! btrfs dev stats -c $mnt >$tmp; then + if ! diff -q $stats_path $tmp; then + mv $stats_path $stats_path.1 + cat $tmp >$stats_path + diff=$(diff -u $stats_path $tmp 2>&1 ||:) + printf "diff of: btrfs dev stats -c %s\n%s\n" "$mnt" "$diff" + exim -odf -i root </dev/null ||: - continue - fi - fi if $check; then + if ! $idle; then + if $dryrun; then + echo "$0: not idle. if this wasnt a dry run, btrfs scrub cancel $mnt" + else + btrfs scrub cancel $mnt &>/dev/null ||: + fi + fi continue fi @@ -144,24 +201,30 @@ main() { sed -rn 's/^\s*scrub started at (.*) and finished.*/\1/p' ) fi - if [[ $date ]]; then + if ! $force && [[ $date ]]; then if $dryrun; then echo "$0: last scrub finish for $mnt: $date" fi date=$(date --date="$date" +%s) - # if date is sooner than 90 days ago + # if date is sooner than 60 days ago # the wiki recommends 30 days or so, but - # it makes the comp lag like shit for a day, - # so I'm going with 90 days. - if (( date > EPOCHSECONDS - 60*60*24*30 )); then + # I'm going with 60 days. + if (( date > EPOCHSECONDS - 60*60*24*60 )); then if $dryrun; then echo "$0: skiping scrub of $mnt, last was $(( (EPOCHSECONDS - date) / 60/60/24 )) days ago, < 30 days" fi continue fi fi - # -c 2 -n 4 is from btrfsmaintenance, does ionice - e btrfs scrub start -Bd -c 2 -n 4 $mnt + # btrfsmaintenance does -c 2 -n 4, but I want lowest pri. + e btrfs scrub start -Bd -c 3 $mnt + + # We normally only do one disk since this is meant to be run in + # downtime and if we try to do all disks, we invariably end up doing + # a scrub after downtime. So, just do one disk per day. + if ! $force; then + return 0 + fi done }