X-Git-Url: https://iankelling.org/git/?a=blobdiff_plain;f=btrbk-run;h=602422987946e1c87b4dbf3e86094108056706c2;hb=af37f71b1c4e1dfa9ebe9a270f4b839cd742cad5;hp=32a0c1d25a176b487d7ec5a13e19f92e01b1783a;hpb=7cd00bca2cb4661c6fd2bb3196f07e54eaea5020;p=distro-setup

diff --git a/btrbk-run b/btrbk-run
index 32a0c1d..6024229 100755
--- a/btrbk-run
+++ b/btrbk-run
@@ -1,88 +1,323 @@
-#!/bin/bash -l
+#!/bin/bash
+# Copyright (C) 2016 Ian Kelling
+
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+
+#     http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+
+# 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.
 
 set -eE -o pipefail
 trap 'echo "$0:$LINENO:error: \"$BASH_COMMAND\" returned $?" >&2' ERR
 
 [[ $EUID == 0 ]] || exec sudo -E "$BASH_SOURCE" "$@"
 
+usage() {
+  cat <<'EOF'
+btrbk-run [OPTIONS]
+usually -t TARGET_HOST
+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"))
+
+# note q is owned by root:1000
+# note p/m is owned 1000:1000 and chmod 700
+
+
+mountpoints=()
+
+rsync_mountpoint=/q
+
+# default options
 conf_only=false
-if [[ $1 == -c ]]; then
-    # mostly for testing
-    conf_only=true
+dry_run=false # mostly for testing
+rate_limit=no
+verbose=true; verbose_arg=-v
+progress_arg="--progress"
+
+default_args_file=/etc/btrbk-run.conf
+if [[ -s $default_args_file ]]; then
+  set -- $(< $default_args_file) "$@"
+fi
+
+temp=$(getopt -l help cl:m:nps:t:vh "$@") || usage 1
+eval set -- "$temp"
+while true; do
+  case $1 in
+    # only creates the config file, does not run btrbk
+    -c) conf_only=true; 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 ;;
+    -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 ;;
+    *) echo "$0: Internal error!" ; exit 1 ;;
+  esac
+done
+
+# usefull commands are resume and archive
+cmd_arg=${1:-run}
+
+if [[ -s $default_args_file ]]; then
+  echo "$0: warning: default btrbk-run options set in $default_args_file (sleeping 5 seconds):"
+  cat $default_args_file
+  sleep 5
+fi
+
+if [[ -v targets && $source ]]; then
+  echo "$0: error: -t and -s are mutually exclusive" >&2
+  exit 1
+fi
+
+echo -e "$0: options: conf_only=$conf_only\ndry_run=$dry_run\nrate_limit=$rate_limit\nverbose=$verbose\ncmd_arg=$cmd_arg"
+
+# set default targets
+if [[ ! -v targets && ! $source ]]; then
+  case $HOSTNAME in
+    x2|kw)
+      if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
+        targets=($HOME_DOMAIN)
+      fi
+      ;;
+    tp)
+      targets=(frodo)
+      if [[ $HOSTNAME == "$MAIL_HOST" ]]; then
+        if timeout -s 9 10 ssh x2 :; then
+          targets+=(x2)
+        fi
+      fi
+      ;;
+    frodo)
+      targets=()
+      ;;
+    *)
+      echo "$0: error: no default targets for this host, use -t"
+      exit 1
+      ;;
+  esac
+fi
+
+if [[ -v targets ]]; then
+  echo "targets: ${targets[*]}"
+fi
+
+if [[ $source ]]; then
+  echo "source: $source"
 fi
 
-target_host=frodo
-cat >/etc/btrbk.conf <<'EOF'
-ssh_identity /root/.ssh/id_rsa
-transaction_syslog daemon
+
+
+# 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 [[ $HOSTNAME == "$MAIL_HOST" ]]; then
+      prospective_mps+=(/o)
+    fi
+    ;;
+esac
+case ${targets[0]} in
+  kw|kww)
+    prospective_mps=(/a)
+    ;;
+esac
+
+for mp in ${prospective_mps[@]}; do # default mountpoints to sync
+  if [[ -e /nocow/btrfs-stale/$mp ]]; then
+    echo "$0: 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
+
+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@$host:/"
+}
+
+
+m() { printf "%s: %s\n" "${0##*/}" "$*";  "$@"; }
+
+
+if ! which btrbk &>/dev/null; then
+  echo "$0: error: no btrbk binary found"
+  exit 1
+fi
+# if our mountpoints are from stale snapshots,
+# it doesn't make sense to do a backup.
+check-subvol-stale ${mountpoints[@]} || exit 1
+
+
+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.
+# The man does not mention a way for them to be together, but
+# I dunno if setting a log level like warn might also output
+# transaction info.
+transaction_syslog local7
+
+# 20%ish speedup[]
+stream_buffer 512m
 
 # so we only run one at a time
 lockfile                   /var/lock/btrbk.lock
 
 # default format of short does not accomidate hourly preservation setting
 timestamp_format long-iso
-# background on timezones. with short/long, timestamps use local time.
-# for long, if your local time moves backwards, by moving timezones or
-# for an hour when daylight savings changes it, you will temporarily get
-# a more aggressive retention policy for the overlapping period, and
-# vice versa for the opposite timezone move. The alternative is using
-# long-iso, which puts timezone info into the timestamp, which means
-# that instead of shifting time, you shift the start of day/week/month
-# which is used for retention to your new local time, which means for
-# example, if you moved forward by 8 hours, the daily/weekly/monthly
-# retention will be 8 hours more aggressive since midnight is at a new
-# time, unless you fake the timzeone using the TZ env variable.
-# However, in the short term, there will be no inconsistencies.
-# I don't see any problem with shifting when the day starts for
-# retention, so I'm using long-iso.
 
 # only make a snapshot if things have changed
 snapshot_create onchange
-# much less snapshots because I have less space on the
-# local filesystem.
-snapshot_preserve 2h 2d
+# I could make this different from target_preserve,
+# if one disk had less space.
+# for now, keeping them equal.
+snapshot_preserve 36h 14d 8w 24m
+snapshot_preserve_min 4h
+snapshot_dir btrbk
 
 # so, total backups = ~89
-target_preserve 48h 14d 8w 24m
-target_preserve_min 6h
+target_preserve 36h 14d 8w 24m
+target_preserve_min 4h
 
 # if something fails and it's not obvious, try doing
 # btrbk -l debug -v dryrun
+
+rate_limit $rate_limit
 EOF
 
 
-case $HOSTNAME in
-    tp|x2)
-        if ! timeout -s 9 10 ssh frodo :; then
-            target_host=$HOME_DOMAIN
-            cat >>/etc/btrbk.conf <<EOF
-ssh_port 2222
-EOF
-        fi
-        ;;
-esac
 
-if [[ $HOSTNAME != frodo ]]; then
-    remote_target="target send-receive ssh://${target_host}/mnt/root/${HOSTNAME}-btrbk"
-fi
+for tg in ${targets[@]:-$HOSTNAME}; do
+  # for an initial run, btrbk requires the dir to exist.
+  ssh root@$tg mkdir -p /mnt/root/btrbk
+done
+
 
-target-section() {
-    root=$1
-    subvol=$2
-    mountpoint $root &>/dev/null || return
-    cat >>/etc/btrbk.conf <<EOF
-volume $root
-subvolume $subvol
-$remote_target
 
+vol=/mnt/root
+for m in ${mountpoints[@]}; do
+  sub=${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
+
+
+# todo: umount first to ensure we don't have any errors
+# todo: do some kill fuser stuff to make umount more reliable
 
-target-section /mnt/iroot i
-target-section /mnt/root q
 
 if $conf_only; then
-    exit
+  exit
 fi
 
-btrbk -q run
+
+
+if $dry_run; then
+  m btrbk -v -n $cmd_arg
+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
+  exit 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
+      li|lk)
+        for x in /p/c/machine_specific/*.hosts; do
+          if grep -qxF $tg $x; then
+            dir=${x%.hosts}
+            rsync-dirs $tg $dir
+          fi
+        done
+        ;;
+    esac
+  done
+fi
+
+PATH=$script_dir:$PATH
+if [[ $source ]]; then
+  m mount-latest-subvol
+else
+  m mount-latest-remote ${targets[@]}
+fi
+
+# todo: move variable data we don't care about backing up
+# to /nocow and symlink it.
+
+
+# background on btrbk timezones. with short/long, timestamps use local time.
+# for long, if your local time moves backwards, by moving timezones or
+# for an hour when daylight savings changes it, you will temporarily get
+# a more aggressive retention policy for the overlapping period, and
+# vice versa for the opposite timezone move. The alternative is using
+# long-iso, which puts timezone info into the timestamp, which means
+# that instead of shifting time, you shift the start of day/week/month
+# which is used for retention to your new local time, which means for
+# example, if you moved forward by 8 hours, the daily/weekly/monthly
+# retention will be 8 hours more aggressive since midnight is at a new
+# time, unless you fake the timzeone using the TZ env variable.
+# However, in the short term, there will be no inconsistencies.
+# I don't see any problem with shifting when the day starts for
+# retention, so I'm using long-iso.
+
+# note to create a long-iso timestamp: date +%Y%m%dT%H%M%S%z