X-Git-Url: https://iankelling.org/git/?p=distro-setup;a=blobdiff_plain;f=btrbk-run;h=68363821af2b1f695655805c83fe683d7f8102cc;hp=eb8a7b5db7e9c89ea051095f55399de541d7e31d;hb=77917a8fbf2032a8b2634a1b3de0879ec45cf213;hpb=ebd0ea323de3eacc11e1a624f8cc4f961d8bdb6c

diff --git a/btrbk-run b/btrbk-run
old mode 100755
new mode 100644
index eb8a7b5..6836382
--- a/btrbk-run
+++ b/btrbk-run
@@ -13,138 +13,332 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-set -eE -o pipefail
-trap 'echo "$0:$LINENO:error: \"$BASH_COMMAND\" returned $?" >&2' ERR
 
-[[ $EUID == 0 ]] || exec sudo -E "$BASH_SOURCE" "$@"
+# todo: if we cancel in the middle of a btrfs send, then run again immediately, the received subvolume doesn't get a Received UUID: field, and we won't mount it. Need to figure out a solution that will fix this.
+
+
+[[ $EUID == 0 ]] || exec sudo -E "${BASH_SOURCE[0]}" "$@"
+
+source /usr/local/lib/err
 
 usage() {
-    echo "top of script file:"
-    sed -n '1,/^[# ]*end command line/{p;b};q' "$0"
-    exit $1
+  cat <<'EOF'
+btrbk-run [OPTIONS]
+usually -t TARGET_HOST or -s SOURCE_HOST
+
+Note, at source location, intentionally not executable, run and read
+install-my-scripts.
+
+EOF
+  echo "top of script file:"
+  sed -n '1,/^[# ]*end command line/{p;b};q' "$0"
+  exit $1
 }
 
-script_dir=$(dirname $(readlink -f "$BASH_SOURCE"))
+rsync-dirs() {
+  local host=$1
+  local path=$2
+  m rsync $dry_run_arg -ahi --relative --delete "$path" "root@$host:/"
+}
 
-# note q is owned by root:1000
-# note p/m is owned 1000:1000 and chmod 700
 
+pre="${0##*/}: "
+m() { if $verbose; then printf "$pre%s\n"  "$*"; fi;  "$@"; }
+e() { printf "$pre%s\n"  "$*"; }
+die() { printf "$pre%s\n" "$*" >&2; echo "exiting with status 1" >&2; exit 1; }
+mexit() { echo "$pre: exiting with status $1"; exit $1; }
+
+# latest $MAIL_HOST
+if [[ -e /b/bash_unpublished/source-state ]]; then
+  source /b/bash_unpublished/source-state
+fi
+
+# note q is owned by root:1000
 
-targets=()
 mountpoints=()
 
 rsync_mountpoint=/q
 
+ret=0
+# default options
 conf_only=false
 dry_run=false # mostly for testing
-resume_arg=
+rate_limit=no
+verbose=true; verbose_arg=-v
+progress_arg="--progress"
+incremental_strict=true
+pull_reexec=false
 
 default_args_file=/etc/btrbk-run.conf
 if [[ -s $default_args_file ]]; then
-    echo "$0: warning: options file default options set:"
-    cat $default_args_file
-    sleep 5
-    set -- $(< $default_args_file) "$@"
+  set -- $(< $default_args_file) "$@"
+  # i havent used this feature yet, so warn about it
+  echo "$0: warning: default btrbk-run options set in $default_args_file (sleeping 5 seconds):"
+  cat $default_args_file
+  sleep 5
 fi
 
-temp=$(getopt -l help hcm:nprt: "$@") || usage 1
+cron=false
+orig_args=("$@")
+temp=$(getopt -l cron,pull-reexec,help cil:m:npqs:t:vh "$@") || usage 1
 eval set -- "$temp"
 while true; do
-    case $1 in
-        -c) conf_only=true; shift ;;
-        -m) IFS=, mountpoints=($2); unset IFS; shift 2 ;;
-        -n) dry_run=true; dry_run_arg=-n; shift ;;
-        -p) progress_arg="--progress"; shift ;;
-        # btrbk arg: Resume only. Skips snapshot creation.
-        -r) resume_arg=-r; shift ;;
-        -t) IFS=, targets=($2); unset IFS; shift 2 ;;
-        -h|--help) usage ;;
-        --) shift; break ;;
-        *) echo "$0: Internal error!" ; exit 1 ;;
-    esac
+  case $1 in
+    --cron)
+      cron=true
+      pre=
+      shift
+      ;;
+    # only creates the config file, does not run btrbk
+    -c) conf_only=true; shift ;;
+    -i) incremental_strict=false; shift ;;
+    # bytes per second, suffix k m g
+    -l) rate_limit=$2; shift 2 ;;
+    # Comma separated mountpoints to backup. This has defaults set below.
+    -m) IFS=, mountpoints=($2); unset IFS; shift 2 ;;
+    -n) dry_run=true; dry_run_arg=-n; shift ;;
+    -p) progress_arg="--progress"; shift ;;
+    --pull-reexec) pull_reexec=true; shift ;;
+    -q) verbose=false; verbose_arg=; progress_arg=; shift ;;
+    # source host to receive a backup from
+    -s) source=$2; shift 2 ;;
+    # target hosts to send to. empty is valid for just doing local
+    # snapshot. we have default hosts we will populate.
+    -t) IFS=, targets=($2); unset IFS; shift 2 ;;
+    -v) verbose=true; verbose_arg=-v; shift ;;
+    -h|--help) usage ;;
+    --) shift; break ;;
+    *) die "Internal error!" ;;
+  esac
 done
 
-echo "$0: options: conf_only=$conf_only, dry_run=$dry_run, resume_arg=$resume_arg"
+# only tested commands are resume and archive
+cmd_arg=${1:-run}
+
+case $cmd_arg in
+  run|resume|archive) : ;;
+  *) die "untested command arg" ;;
+esac
+
+if (( $# > 1 )); then
+  die: "only 1 nonoption arg is supported"
+fi
+
+if [[ -v targets && $source ]]; then
+  die "error: -t and -s are mutually exclusive"
+fi
+
+if $verbose; then
+  printf "options: conf_only=%s\ndry_run=%s\nrate_limit=%s\nverbose=%s\ncmd_arg=%s" "$conf_only" "$dry_run" "$rate_limit" "$verbose" "$cmd_arg"
+fi
+### end options parsing
+
+if ! btrbk --version |& grep 0.29.0 >/dev/null; then
+  cd /a/opt/btrbk
+  m sudo make install
+fi
+
+# TODO: i wonder if there should be an option to send to the default
+# targets, plus any given on the command line.
 
 # set default targets
-if ! (( ${#targets[@]} )); then
-    case $HOSTNAME in
-        x2)
-            if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
-                targets=($HOME_DOMAIN)
-            fi
-            ;;
-        treetowl)
-            targets=(frodo)
-            if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
-                if timeout -s 9 10 ssh x2 :; then
-                    targets+=(x2)
-                fi
-            fi
-            ;;
-        *)
-            echo "$0: error: no default targets for this host, use -t"
-            exit 1
-            ;;
-    esac
+if [[ ! -v targets && ! $source ]]; then
+  if [[ $HOSTNAME != "$MAIL_HOST" ]] && $cron ; then
+    echo "MAIL_HOST=$MAIL_HOST, nothing to do"
+    mexit 0
+  fi
+
+  at_work=false
+
+  case $HOSTNAME in
+    kw)
+      at_work=true
+      ;;&
+    x2|x3)
+      if ping -q -c1 -w1 hal.office.fsf.org \
+          && ip n show 192.168.0.26 | grep . &>/dev/null; then
+        at_work=true
+      fi
+      ;;&
+    kw|x2|x3)
+      if $at_work && ping -q -c1 -w1 iank.vpn.office.fsf.org &>/dev/null; then
+        home=iank.vpn.office.fsf.org
+      else
+        home=b8.nz
+      fi
+      ;;&
+    kw)
+      targets=($home x3)
+      ;;
+    x2|x3)
+      targets=($home kw)
+      ;;
+    tp)
+      # kd disabled temporarily while its hot and i plan to work on it.
+      #targets=(frodo kd)
+      #targets=(frodo x2.b8.nz)
+      targets=(frodo)
+      # might not be connected to the vpn
+      if timeout -s 9 6 ssh kw :; then
+        targets+=(kw)
+      fi
+      ;;
+    kd)
+      targets=(frodo tp)
+      # might not be connected to the vpn
+      if timeout -s 9 6 ssh kw :; then
+        targets+=(kw)
+      fi
+      ;;
+    *)
+      die "error: no default targets for this host, use -t"
+      ;;
+  esac
 fi
 
-echo "targets: ${targets[*]}"
+if [[ -v targets ]]; then
+  echo "targets: ${targets[*]}"
+fi
 
+if [[ $source ]]; then
+  echo "source: $source"
+fi
 
 
-# set default mountpoints
-if ! (( ${#mountpoints[@]} )); then
-    prospective_mps=(/a /q)
-    if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
-        prospective_mps+=(/o)
+if [[ $mountpoints ]]; then
+  for mp in ${mountpoints[@]}; do # default mountpoints to sync
+    if [[ -e /nocow/btrfs-stale/$mp ]]; then
+      die "error: $mp is stale, mount-latest-subvol first"
     fi
-    for tg in ${targets[@]}; do
-        if [[ $tg == frodo && $HOSTNAME == treetowl ]]; then
-            prospective_mps+=(/i)
+  done
+else
+  # set default mountpoints
+  case $HOSTNAME in
+    # no remote backups atm. note, if we do enable this, configuration below will need some changes.
+    #  frodo)
+    #    prospective_mps=(/i)
+    #    ;;
+    *)
+      prospective_mps=(/a /q)
+      if [[ $source ]]; then
+        source_state="$(ssh $source cat /a/bin/bash_unpublished/source-state)"
+        eval "$source_state"
+        source_host="$(ssh $source cat /etc/hostname)"
+        if [[ $source_host == "$MAIL_HOST" ]]; then
+          prospective_mps+=(/o)
         fi
-    done
-    for mp in ${prospective_mps[@]}; do # default mountpoints to sync
-        if awk '{print $2}' /etc/fstab | grep -xF $mp &>/dev/null; then
-            mountpoints+=($mp)
+      else
+        if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
+          prospective_mps+=(/o)
         fi
-    done
+      fi
+      ;;
+  esac
+  for mp in ${prospective_mps[@]}; do # default mountpoints to sync
+    if [[ -e /nocow/btrfs-stale/$mp ]]; then
+      e "warning: $mp stale, not adding to default mountpoints"
+      continue
+    fi
+    if awk '{print $2}' /etc/fstab | grep -xF $mp &>/dev/null; then
+      mountpoints+=($mp)
+    fi
+  done
 fi
 
 echo "mountpoints: ${mountpoints[*]}"
 
 ##### end command line parsing ########
 
-rsync-dirs() {
-    local host=$1
-    local path=$2
-    m rsync $dry_run_arg -ahi --relative --delete "$path" "root@$host:/"
-}
+if [[ $source ]]; then
+  if [[ $(ssh $source systemctl is-active btrbk.service) != inactive ]]; then
+    echo "$0: error: cron btrbk is running on source. exiting out of caution"
+    mexit 1
+  fi
+fi
+
+# pull_reexec stops us from getting into an infinite loop if there is some
+# kind of weird problem
+pulla=false
+for m in "${mountpoints[@]}"; do
+  if [[ $m == /a ]]; then
+    pulla=true
+    break
+  fi
+done
+if ! $pull_reexec && [[ $source ]] && $pulla ; then
+  tmpf=$(mktemp)
+  scp $source:/a/bin/distro-setup/btrbk-run $tmpf
+  if ! diff -q $tmpf $BASH_SOURCE; then
+    e "found newer version on host $source. reexecing"
+    install -T $tmpf /usr/local/bin/btrbk-run
+    m /usr/local/bin/btrbk-run --pull-reexec "${orig_args[@]}"
+    mexit 0
+  fi
+fi
+
 
-vol-conf() {
-    cat >>/etc/btrbk.conf <<EOF
-volume $vol
-EOF
-}
-sub-conf() {
-    cat >>/etc/btrbk.conf <<EOF
-subvolume $sub
-EOF
-}
-tg-conf() {
-    cat >>/etc/btrbk.conf <<EOF
-target send-receive ssh://$tg$vol/btrbk
-EOF
-}
-m() { printf "%s: %s\n" "${0##*/}" "$*";  "$@"; }
 
 
 if ! which btrbk &>/dev/null; then
-    echo "$0: error: no btrbk binary found"
+  die "error: no btrbk binary found"
 fi
+# if our mountpoints are from stale snapshots,
+# it doesn't make sense to do a backup.
+check-subvol-stale ${mountpoints[@]} || die "found stale mountpoints in ${mountpoints[*]}"
 
-cat >/etc/btrbk.conf <<'EOF'
-ssh_identity /root/.ssh/id_rsa
+# for an initial run, btrbk requires the dir to exist.
+mkdir -p /mnt/root/btrbk
+local_zone=$(date +%z)
+
+if [[ $source ]]; then
+  if ! zone=$(ssh root@$source date +%z); then
+    die failed to ssh to root@$source
+  fi
+  if [[ $zone != $local_zone ]]; then
+    die "error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
+  fi
+
+else
+
+  sshable=()
+  sshfail=()
+  min_idle_ms=$((1000 * 60 * 15))
+  for h in ${targets[@]}; do
+    if zone=$(ssh root@$h "mkdir -p /mnt/root/btrbk && date +%z"); then
+      # This is a separate ssh because xprintidle can fail and thats ok.
+      if $cron && idle_ms=$(timeout -s 9 6 ssh $h DISPLAY=:0 xprintidle); then
+        if (( idle_ms < min_idle_ms )); then
+
+          # Ignore this host. i sometimes use a non-main machine for
+          # testing or web browsing, knowing that everything will be wiped
+          # by the next backup, but I dont want it to happen as Im using
+          # it from cronjob.
+          e "warning: $h: active X session in the last 15 minutes, skipping for now"
+          continue
+        fi
+      fi
+      sshable+=($h)
+      if [[ $zone != $local_zone ]]; then
+        die "error: dont confuse yourself with multiple time zones. $h has different timezone than localhost"
+      fi
+    else
+      sshfail+=($h)
+    fi
+  done
+  if [[ ! ${sshable[@]} ]] || { ! $cron && [[ ${sshfail[@]} ]]; }; then
+    die "failed to ssh to hosts: ${sshfail[*]}"
+  else
+    if [[ ${sshfail[@]} ]]; then
+      ret=1
+      e "error: failed to ssh to ${sshfail[*]} but continuing with other hosts"
+    fi
+    targets=(${sshable[@]})
+  fi
+fi
+
+
+cat >/etc/btrbk.conf <<EOF
+ssh_identity /root/.ssh/home
 # Just a guess that local7 is a good facility to pick.
 # It's a bit odd that the transaction log has to be logged to
 # a file or syslog, while other output is sent to std out.
@@ -153,6 +347,10 @@ ssh_identity /root/.ssh/id_rsa
 # transaction info.
 transaction_syslog local7
 
+# note, i had this because man said 20% speedup, but ran into
+# this issue, https://github.com/digint/btrbk/issues/275
+#stream_buffer 512m
+
 # so we only run one at a time
 lockfile                   /var/lock/btrbk.lock
 
@@ -174,38 +372,36 @@ target_preserve_min 4h
 
 # if something fails and it's not obvious, try doing
 # btrbk -l debug -v dryrun
-EOF
-
-
-# if our mountpoints are from stale snapshots,
-# it doesn't make sense to do a backup.
-check-subvol-stale ${mountpoints[@]} || exit 1
 
-for tg in ${targets[@]}; do
-    # for an initial run, btrbk requires the dir to exist
-    ssh root@$tg mkdir -p /mnt/root/btrbk
-done
+rate_limit $rate_limit
+EOF
 
+if $incremental_strict; then
+  cat >>/etc/btrbk.conf <<EOF
+incremental strict
+EOF
+fi
 
+vol=/mnt/root
 for m in ${mountpoints[@]}; do
-    # for /i, some special cases. there is just one static target and direction.
-    if [[ $m == /i ]]; then
-        vol=/mnt/iroot
-        vol-conf
-        sub=i
-        sub-conf
-        tg=frodo
-        vol=/mnt/root
-        tg-conf
-    else
-        vol=/mnt/root
-        vol-conf
-        sub=${m##*/}
-        sub-conf
-        for tg in ${targets[@]}; do
-            tg-conf
-        done
-    fi
+  sub=${m##*/}
+  if [[ $source ]]; then
+    cat >>/etc/btrbk.conf <<EOF
+volume ssh://$source$vol
+subvolume $sub
+target send-receive $vol/btrbk
+EOF
+  else
+    cat >>/etc/btrbk.conf <<EOF
+volume $vol
+subvolume $sub
+EOF
+    for tg in ${targets[@]}; do
+      cat >>/etc/btrbk.conf <<EOF
+target send-receive ssh://$tg$vol/btrbk
+EOF
+    done
+  fi
 done
 
 
@@ -214,37 +410,51 @@ done
 
 
 if $conf_only; then
-    exit
+  mexit 0
 fi
 
+
+
 if $dry_run; then
-    m btrbk -n $resume_arg run
-else
-    # -q and just using the syslog option seemed nice,
-    # but it doesn't show when a send has a parent and when it doesn't.
-    m btrbk $progress_arg $resume_arg run
+  m btrbk -v -n $cmd_arg
+  mexit 0
+elif [[ $cmd_arg == archive ]]; then
+  if [[ $source ]]; then
+    m btrbk $verbose_arg $progress_arg $cmd_arg ssh://$source$vol $vol
+  else
+    for tg in ${targets[@]}; do
+      m btrbk $verbose_arg $progress_arg $cmd_arg $vol ssh://$tg$vol
+    done
+  fi
+  mexit 0
 fi
+# -q and just using the syslog option seemed nice,
+# but it doesn't show when a send has a parent and when it doesn't.
+m btrbk $verbose_arg $progress_arg $cmd_arg
 
 # if we have it, sync to systems which don't
 if mountpoint $rsync_mountpoint >/dev/null; then
-    for tg in ${targets[@]}; do
-        case $tg in
-            tp|li|lk)
-                for x in /p/c/machine_specific/*.hosts; do
-                    if grep -qxF $tg $x; then
-                        dir=${x%.hosts}
-                        rsync-dirs $tg $dir
-                    fi
-                done
-                ;;
-        esac
-    done
+  for tg in ${targets[@]}; do
+    case $tg in
+      li|lk)
+        for x in /p/c/machine_specific/*.hosts; do
+          if grep -qxF $tg $x; then
+            dir=${x%.hosts}
+            rsync-dirs $tg $dir
+          fi
+        done
+        ;;
+    esac
+  done
 fi
 
-if ! $dry_run; then
-    m $script_dir/mount-latest-remote ${targets[@]}
+if [[ $source ]]; then
+  m mount-latest-subvol
+else
+  m /a/exe/mount-latest-remote ${targets[@]}
 fi
 
+mexit $ret
 
 # todo: move variable data we don't care about backing up
 # to /nocow and symlink it.