ccollect/ccollect

866 lines
21 KiB
Bash
Executable file

#!/bin/sh
#
# 2005-2013 Nico Schottelius (nico-ccollect at schottelius.org)
# 2016-2019 Darko Poljak (darko.poljak at gmail.com)
#
# This file is part of ccollect.
#
# ccollect is free software: you can redistribute it and/or modify
# it under the terms of the GNU General Public License as published by
# the Free Software Foundation, either version 3 of the License, or
# (at your option) any later version.
#
# ccollect is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU General Public License for more details.
#
# You should have received a copy of the GNU General Public License
# along with ccollect. If not, see <http://www.gnu.org/licenses/>.
#
# Initially written for SyGroup (www.sygroup.ch)
# Date: Mon Nov 14 11:45:11 CET 2005
# Error upon expanding unset variables:
set -u
#
# Standard variables (stolen from cconf)
#
__pwd="$(pwd -P)"
__mydir="${0%/*}"; __abs_mydir="$(cd "$__mydir" && pwd -P)"
__myname=${0##*/}; __abs_myname="$__abs_mydir/$__myname"
#
# where to find our configuration and temporary file
#
CCOLLECT_CONF="${CCOLLECT_CONF:-/etc/ccollect}"
CSOURCES="${CCOLLECT_CONF}/sources"
CDEFAULTS="${CCOLLECT_CONF}/defaults"
CPREEXEC="${CDEFAULTS}/pre_exec"
CPOSTEXEC="${CDEFAULTS}/post_exec"
CMARKER=".ccollect-marker"
export TMP="$(mktemp "/tmp/${__myname}.XXXXXX")"
CONTROL_PIPE="/tmp/${__myname}-control-pipe"
VERSION="2.5"
RELEASE="2019-05-01"
HALF_VERSION="ccollect ${VERSION}"
FULL_VERSION="ccollect ${VERSION} (${RELEASE})"
#
# CDATE: how we use it for naming of the archives
# DDATE: how the user should see it in our output (DISPLAY)
#
CDATE="date +%Y%m%d-%H%M"
DDATE="date +%Y-%m-%d-%H:%M:%S"
SDATE="date +%s"
#
# LOCKING: use flock if available, otherwise mkdir
# Locking is done for each source so that only one instance per source
# can run.
#
# Use CCOLLECT_CONF directory for lock files.
# This directory can be set arbitrary so it is writable for user
# executing ccollect.
LOCKDIR="${CCOLLECT_CONF}"
# printf pattern: ccollect_<source>.lock
LOCKFILE_PATTERN="ccollect_%s.lock"
LOCKFD=4
#
# locking functions using flock
#
lock_flock()
{
# $1 = source to backup
lockfile="${LOCKDIR}/$(printf "${LOCKFILE_PATTERN}" "$1")"
eval "exec ${LOCKFD}> ${lockfile}"
flock -n ${LOCKFD} && return 0 || return 1
}
unlock_flock()
{
# $1 = source to backup
lockfile="${LOCKDIR}/$(printf "${LOCKFILE_PATTERN}" "$1")"
eval "exec ${LOCKFD}>&-"
rm -f "${lockfile}"
}
#
# locking functions using mkdir (mkdir is atomic)
#
lock_mkdir()
{
# $1 = source to backup
lockfile="${LOCKDIR}/$(printf "${LOCKFILE_PATTERN}" "$1")"
mkdir "${lockfile}" && return 0 || return 1
}
unlock_mkdir()
{
# $1 = source to backup
lockfile="${LOCKDIR}/$(printf "${LOCKFILE_PATTERN}" "$1")"
rmdir "${lockfile}"
}
#
# determine locking tool: flock or mkdir
#
if $(which flock > /dev/null 2>&1)
then
lockf="lock_flock"
unlockf="unlock_flock"
else
lockf="lock_mkdir"
unlockf="unlock_mkdir"
fi
#
# unset values
#
PARALLEL=""
MAX_JOBS=""
USE_ALL=""
LOGFILE=""
SYSLOG=""
# e - only errors, a - all output
LOGLEVEL="a"
LOGONLYERRORS=""
#
# catch signals
#
TRAPFUNC="rm -f \"${TMP}\""
trap "${TRAPFUNC}" 1 2 15
#
# Functions
#
# check if we are running interactive or non-interactive
# see: http://www.tldp.org/LDP/abs/html/intandnonint.html
_is_interactive()
{
[ -t 0 -o -p /dev/stdin ]
}
#
# ssh-"feature": we cannot do '... read ...; ssh ...; < file',
# because ssh reads stdin! -n does not work -> does not ask for password
# Also allow deletion for files without the given suffix
#
delete_from_file()
{
file="$1"; shift
suffix="" # It will be set, if deleting incomplete backups.
[ $# -eq 1 ] && suffix="$1" && shift
# dirs for deletion will be moved to this trash dir inside destination dir
# - for fast mv operation
trash="$(mktemp -d ".trash.XXXXXX")"
while read to_remove; do
mv "${to_remove}" "${trash}" ||
_exit_err "Moving ${to_remove} to ${trash} failed."
set -- "$@" "${to_remove}"
if [ "${suffix}" ]; then
to_remove_no_suffix="$(echo ${to_remove} | sed "s/$suffix\$//")"
mv "${to_remove_no_suffix}" "${trash}" ||
_exit_err "Moving ${to_remove_no_suffix} to ${trash} failed."
set -- "$@" "${to_remove_no_suffix}"
fi
done < "${file}"
_techo "Removing $@ in ${trash}..."
empty_dir=".empty-dir"
mkdir "${empty_dir}" || _exit_err "Empty directory ${empty_dir} cannot be created."
[ "${VVERBOSE}" ] && echo "Starting: rsync -a --delete ${empty_dir} ${trash}"
# rsync needs ending slash for directory content
rsync -a --delete "${empty_dir}/" "${trash}/" || _exit_err "Removing $@ failed."
rmdir "${trash}" || _exit_err "Removing ${trash} directory failed"
rmdir "${empty_dir}" || _exit_err "Removing ${empty_dir} directory failed"
_techo "Removing $@ in ${trash} finished."
}
display_version()
{
echo "${FULL_VERSION}"
exit 0
}
usage()
{
cat << eof
${__myname}: [args] <interval name> <sources to backup>
ccollect creates (pseudo) incremental backups
-h, --help: Show this help screen
-a, --all: Backup all sources specified in ${CSOURCES}
-e, --errors: Log only errors
-j [max], --jobs [max] Specifies the number of jobs to run simultaneously.
If max is not specified then parallelise all jobs.
-l FILE, --logfile FILE Log to specified file
-p, --parallel: Parallelise backup processes (deprecated from 2.0)
-s, --syslog: Log to syslog with tag ccollect
-v, --verbose: Be very verbose (uses set -x)
-V, --version: Print version information
This is version ${VERSION} released on ${RELEASE}.
Retrieve latest ccollect at http://www.nico.schottelius.org/software/ccollect/
eof
exit 0
}
# locking functions
lock()
{
"${lockf}" "$@" || _exit_err \
"Only one instance of ${__myname} for source \"$1\" can run at one time."
}
unlock()
{
"${unlockf}" "$@"
}
# time displaying echo
# stdout version
_techo_stdout()
{
echo "$(${DDATE}): $@"
}
# syslog version
_techo_syslog()
{
logger -t ccollect "$@"
}
# specified file version
_techo_file()
{
_techo_stdout "$@" >> "${LOGFILE}"
}
# determine _techo version before parsing options
if _is_interactive
then
_techof="_techo_stdout"
else
_techof="_techo_syslog"
fi
# _techo with determined _techo version
_techo()
{
if [ "${LOGLEVEL}" = "a" ]
then
set -- ${name:+"[${name}]"} $@
"${_techof}" "$@"
fi
}
_techo_err()
{
_techo "Error: $@"
}
_exit_err()
{
_techo_err "$@"
rm -f "${TMP}"
exit 1
}
#
# Parse options
#
while [ "$#" -ge 1 ]; do
case "$1" in
-a|--all)
USE_ALL=1
;;
-p|--parallel)
_techo "Warning: -p, --parallel option is deprecated," \
"use -j, --jobs instead."
PARALLEL=1
MAX_JOBS=""
;;
-j|--jobs)
PARALLEL=1
if [ "$#" -ge 2 ]
then
case "$2" in
-*)
;;
*)
MAX_JOBS=$2
shift
;;
esac
fi
;;
-e|--errors)
LOGONLYERRORS="1"
;;
-l|--logfile)
if [ "$#" -ge 2 ]
then
case "$2" in
-*)
_exit_err "Missing log file"
;;
*)
LOGFILE="$2"
shift
;;
esac
else
_exit_err "Missing log file"
fi
;;
-s|--syslog)
SYSLOG="1"
;;
-v|--verbose)
set -x
;;
-V|--version)
display_version
;;
--)
# ignore the -- itself
shift
break
;;
-h|--help|-*)
usage
;;
*)
break
;;
esac
shift
done
# determine _techo version and logging level after parsing options
if [ "${LOGFILE}" ]
then
_techof="_techo_file"
LOGLEVEL="a"
elif _is_interactive
then
if [ "${SYSLOG}" ]
then
_techof="_techo_syslog"
LOGLEVEL="a"
else
_techof="_techo_stdout"
LOGLEVEL="e"
fi
else
_techof="_techo_syslog"
LOGLEVEL="a"
fi
if [ "${LOGFILE}" -o "${SYSLOG}" ]
then
if [ "${LOGONLYERRORS}" ]
then
LOGLEVEL="e"
fi
fi
# check that MAX_JOBS is natural number > 0
# empty string means run all in parallel
echo "${MAX_JOBS}" | grep -q -E '^[1-9][0-9]*$|^$'
if [ "$?" -ne 0 ]
then
_exit_err "Invalid max jobs value \"${MAX_JOBS}\""
fi
#
# Setup interval
#
if [ $# -ge 1 ]; then
export INTERVAL="$1"
shift
else
usage
fi
#
# Check for configuraton directory
#
[ -d "${CCOLLECT_CONF}" ] || _exit_err "No configuration found in " \
"\"${CCOLLECT_CONF}\" (is \$CCOLLECT_CONF properly set?)"
#
# Create (portable!) source "array"
#
export no_sources=0
if [ "${USE_ALL}" = 1 ]; then
#
# Get sources from source configuration
#
( cd "${CSOURCES}" && ls -1 > "${TMP}" ) || \
_exit_err "Listing of sources failed. Aborting."
while read tmp; do
eval export source_${no_sources}=\"${tmp}\"
no_sources=$((${no_sources}+1))
done < "${TMP}"
else
#
# Get sources from command line
#
while [ "$#" -ge 1 ]; do
eval arg=\"\$1\"; shift
eval export source_${no_sources}=\"${arg}\"
no_sources="$((${no_sources}+1))"
done
fi
#
# Need at least ONE source to backup
#
if [ "${no_sources}" -lt 1 ]; then
usage
else
_techo "${HALF_VERSION}: Beginning backup using interval ${INTERVAL}"
fi
#
# Look for pre-exec command (general)
#
if [ -x "${CPREEXEC}" ]; then
_techo "Executing ${CPREEXEC} ..."
"${CPREEXEC}"; ret=$?
_techo "Finished ${CPREEXEC} (return code: ${ret})."
[ "${ret}" -eq 0 ] || _exit_err "${CPREEXEC} failed. Aborting"
fi
################################################################################
#
# Let's do the backup - here begins the real stuff
#
# in PARALLEL mode:
# * create control pipe
# * determine number of jobs to start at once
if [ "${PARALLEL}" ]; then
mkfifo "${CONTROL_PIPE}"
# fd 5 is tied to control pipe
eval "exec 5<>${CONTROL_PIPE}"
TRAPFUNC="${TRAPFUNC}; rm -f \"${CONTROL_PIPE}\""
trap "${TRAPFUNC}" 0 1 2 15
# determine how much parallel jobs to prestart
if [ "${MAX_JOBS}" ]
then
if [ "${MAX_JOBS}" -le "${no_sources}" ]
then
prestart="${MAX_JOBS}"
else
prestart="${no_sources}"
fi
else
prestart=0
fi
fi
source_no=0
while [ "${source_no}" -lt "${no_sources}" ]; do
#
# Get current source
#
eval export name=\"\$source_${source_no}\"
source_no=$((${source_no}+1))
#
# Start ourself, if we want parallel execution
#
if [ "${PARALLEL}" ]; then
if [ ! "${MAX_JOBS}" ]
then
# run all in parallel
"$0" "${INTERVAL}" "${name}" &
continue
elif [ "${prestart}" -gt 0 ]
then
# run prestart child if pending
{ "$0" "${INTERVAL}" "${name}"; printf '\n' >&5; } &
prestart=$((${prestart} - 1))
continue
else
# each time a child finishes we get a line from the pipe
# and then launch another child
while read line
do
{ "$0" "${INTERVAL}" "${name}"; printf '\n' >&5; } &
# get out of loop so we can contnue with main loop
# for next source
break
done <&5
continue
fi
fi
#
# Start subshell for easy log editing
#
(
backup="${CSOURCES}/${name}"
c_source="${backup}/source"
c_dest="${backup}/destination"
c_pre_exec="${backup}/pre_exec"
c_post_exec="${backup}/post_exec"
#
# Stderr to stdout, so we can produce nice logs
#
exec 2>&1
#
# Record start of backup: internal and for the user
#
begin_s="$(${SDATE})"
_techo "Beginning to backup"
#
# Standard configuration checks
#
if [ ! -e "${backup}" ]; then
_exit_err "Source \"${backup}\" does not exist."
fi
#
# Configuration _must_ be a directory (cconfig style)
#
if [ ! -d "${backup}" ]; then
_exit_err "\"${backup}\" is not a cconfig-directory. Skipping."
fi
#
# Acquire lock for source. If lock cannot be acquired, lock will exit
# with error message.
#
lock "${name}"
# redefine trap to also unlock (rm lockfile)
TRAPFUNC="${TRAPFUNC}; unlock \"${name}\""
trap "${TRAPFUNC}" 1 2 15
#
# First execute pre_exec, which may generate destination or other parameters
#
if [ -x "${c_pre_exec}" ]; then
_techo "Executing ${c_pre_exec} ..."
"${c_pre_exec}"; ret="$?"
_techo "Finished ${c_pre_exec} (return code ${ret})."
[ "${ret}" -eq 0 ] || _exit_err "${c_pre_exec} failed. Skipping."
fi
#
# Read source configuration
#
for opt in verbose very_verbose summary exclude rsync_options \
delete_incomplete rsync_failure_codes \
mtime quiet_if_down ; do
if [ -f "${backup}/${opt}" -o -f "${backup}/no_${opt}" ]; then
eval c_$opt=\"${backup}/$opt\"
else
eval c_$opt=\"${CDEFAULTS}/$opt\"
fi
done
#
# Interval definition: First try source specific, fallback to default
#
c_interval="$(cat "${backup}/intervals/${INTERVAL}" 2>/dev/null)"
if [ -z "${c_interval}" ]; then
c_interval="$(cat "${CDEFAULTS}/intervals/${INTERVAL}" 2>/dev/null)"
if [ -z "${c_interval}" ]; then
_exit_err "No definition for interval \"${INTERVAL}\" found. Skipping."
fi
fi
#
# Sort by ctime (default) or mtime (configuration option)
#
if [ -f "${c_mtime}" ] ; then
TSORT="t"
else
TSORT="tc"
fi
#
# Source configuration checks
#
if [ ! -f "${c_source}" ]; then
_exit_err "Source description \"${c_source}\" is not a file. Skipping."
else
source=$(cat "${c_source}"); ret="$?"
if [ "${ret}" -ne 0 ]; then
_exit_err "Source ${c_source} is not readable. Skipping."
fi
fi
#
# Destination is a path
#
if [ ! -f "${c_dest}" ]; then
_exit_err "Destination ${c_dest} is not a file. Skipping."
else
ddir="$(cat "${c_dest}")"; ret="$?"
if [ "${ret}" -ne 0 ]; then
_exit_err "Destination ${c_dest} is not readable. Skipping."
fi
fi
#
# Parameters: ccollect defaults, configuration options, user options
#
#
# Rsync standard options (archive will be added after is-up-check)
#
set -- "$@" "--delete" "--numeric-ids" "--relative" \
"--delete-excluded" "--sparse"
#
# Exclude list
#
if [ -f "${c_exclude}" ]; then
set -- "$@" "--exclude-from=${c_exclude}"
fi
#
# Output a summary
#
if [ -f "${c_summary}" ]; then
set -- "$@" "--stats"
fi
#
# Verbosity for rsync, rm, and mkdir
#
VVERBOSE=""
if [ -f "${c_very_verbose}" ]; then
set -- "$@" "-vv"
VVERBOSE="-v"
elif [ -f "${c_verbose}" ]; then
set -- "$@" "-v"
fi
#
# Extra options for rsync provided by the user
#
if [ -f "${c_rsync_options}" ]; then
while read line; do
# Trim line.
ln=$(echo "${line}" | awk '{$1=$1;print;}')
# Only if ln is non zero length string.
#
# If ln is empty then rsync '' DEST evaluates
# to transfer current directory to DEST which would
# with specific options destroy DEST content.
if [ -n "${ln}" ]
then
set -- "$@" "${ln}"
fi
done < "${c_rsync_options}"
fi
#
# Check: source is up and accepting connections (before deleting old backups!)
#
if ! rsync "$@" "${source}" >/dev/null 2>"${TMP}" ; then
if [ ! -f "${c_quiet_if_down}" ]; then
cat "${TMP}"
fi
_exit_err "Source ${source} is not readable. Skipping."
fi
#
# Add --archive for real backup (looks nice in front)
#
set -- "--archive" "$@"
#
# Check: destination exists?
#
cd "${ddir}" || _exit_err "Cannot change to ${ddir}. Skipping."
#
# Check incomplete backups (needs echo to remove newlines)
#
ls -1 | grep "${CMARKER}\$" > "${TMP}"; ret=$?
if [ "$ret" -eq 0 ]; then
_techo "Incomplete backups: $(echo $(cat "${TMP}"))"
if [ -f "${c_delete_incomplete}" ]; then
delete_from_file "${TMP}" "${CMARKER}" &
fi
fi
#
# Include current time in name, not the time when we began to remove above
#
export destination_name="${INTERVAL}.$(${CDATE}).$$-${source_no}"
export destination_dir="${ddir}/${destination_name}"
#
# Check: maximum number of backups is reached?
#
count="$(ls -1 | grep -c "^${INTERVAL}\\.")"
_techo "Existing backups: ${count} Total keeping backups: ${c_interval}"
if [ "${count}" -ge "${c_interval}" ]; then
# Use oldest directory as new backup destination directory.
# It need not to be deleted, rsync will sync its content.
oldest_bak=$(ls -${TSORT}1r | grep "^${INTERVAL}\\." | head -n 1 || \
_exit_err "Listing oldest backup failed")
_techo "Using ${oldest_bak} for destination dir"
mv "${oldest_bak}" "${destination_dir}" ||
_exit_err "Moving oldest backup ${oldest_bak} to ${destination_dir} failed."
# Touch dest dir so it is not sorted wrong in listings below.
touch "${destination_dir}"
# We have something to remove only if count > interval.
if [ "${count}" -gt "${c_interval}" ]; then
remove="$((${count} - ${c_interval}))"
_techo "Removing ${remove} backup(s)..."
ls -${TSORT}1r | grep "^${INTERVAL}\\." | head -n "${remove}" > "${TMP}" || \
_exit_err "Listing old backups failed"
delete_from_file "${TMP}" &
fi
fi
#
# Check for backup directory to clone from: Always clone from the latest one!
#
last_dir="$(ls -${TSORT}p1 | grep '/$' | head -n 1)" || \
_exit_err "Failed to list contents of ${ddir}."
#
# Clone from old backup, if existing
#
if [ "${last_dir}" ]; then
set -- "$@" "--link-dest=${ddir}/${last_dir}"
_techo "Hard linking from ${last_dir}"
fi
#
# Mark backup running and go back to original directory
#
touch "${destination_dir}${CMARKER}"
cd "${__abs_mydir}" || _exit_err "Cannot go back to ${__abs_mydir}."
#
# the rsync part
#
_techo "Transferring files..."
rsync "$@" "${source}" "${destination_dir}"; ret=$?
_techo "Finished backup (rsync return code: $ret)."
#
# Set modification time (mtime) to current time, if sorting by mtime is enabled
#
[ -f "$c_mtime" ] && touch "${destination_dir}"
#
# Check if rsync exit code indicates failure.
#
fail=""
if [ -f "$c_rsync_failure_codes" ]; then
while read code ; do
if [ "$ret" = "$code" ]; then
fail=1
fi
done <"${c_rsync_failure_codes}"
fi
#
# Remove marking here unless rsync failed.
#
if [ -z "$fail" ]; then
rm "${destination_dir}${CMARKER}" || \
_exit_err "Removing ${destination_dir}${CMARKER} failed."
if [ "${ret}" -ne 0 ]; then
_techo "Warning: rsync exited non-zero, the backup may be broken (see rsync errors)."
fi
else
_techo "Warning: rsync failed with return code $ret."
fi
#
# post_exec
#
if [ -x "${c_post_exec}" ]; then
_techo "Executing ${c_post_exec} ..."
"${c_post_exec}"; ret=$?
_techo "Finished ${c_post_exec}."
if [ "${ret}" -ne 0 ]; then
_exit_err "${c_post_exec} failed."
fi
fi
#
# Time calculation
#
end_s="$(${SDATE})"
full_seconds="$((${end_s} - ${begin_s}))"
hours="$((${full_seconds} / 3600))"
minutes="$(((${full_seconds} % 3600) / 60))"
seconds="$((${full_seconds} % 60))"
_techo "Backup lasted: ${hours}:${minutes}:${seconds} (h:m:s)"
unlock "${name}"
# wait for children (doing delete_from_file) if any still running
wait
) || exit
done
#
# Be a good parent and wait for our children, if they are running wild parallel
# After all children are finished then remove control pipe.
#
if [ "${PARALLEL}" ]; then
_techo "Waiting for children to complete..."
wait
rm -f "${CONTROL_PIPE}"
fi
#
# Look for post-exec command (general)
#
if [ -x "${CPOSTEXEC}" ]; then
_techo "Executing ${CPOSTEXEC} ..."
"${CPOSTEXEC}"; ret=$?
_techo "Finished ${CPOSTEXEC} (return code: ${ret})."
if [ "${ret}" -ne 0 ]; then
_techo "${CPOSTEXEC} failed."
fi
fi
rm -f "${TMP}"
_techo "Finished"