22

✏️ Editing: 950_copy_result_files.sh

#
# output/default/950_copy_result_files.sh
# Copy the resulting files to the output location.
#

# For example for "rear mkbackuponly" there are usually no result files
# that would need to be copied here to the output location:
test "${RESULT_FILES[*]:-}" || return 0

local scheme=$( url_scheme $OUTPUT_URL )
local host=$( url_host $OUTPUT_URL )
local path=$( url_path $OUTPUT_URL )

if [ -z "$OUTPUT_URL" ] || ! scheme_accepts_files $scheme ; then
    if [ "$scheme" == "null" -o -z "$OUTPUT_URL" ] ; then
        # There are result files to copy, but OUTPUT_URL=null indicates that we are not interested in them
        # TODO: empty OUTPUT_URL seems to be equivalent to null, should we continue to allow that,
        # or enforce setting it explicitly?
        return 0
    else
        # There are files to copy, but schemes like tape: do not allow files to be stored. The files would be lost.
        # Do not allow that.
        # Schemes like obdr: that store the results themselves should clear RESULT_FILES to indicate that nothing is to be done.
        # Is this considered a bug in ReaR (BugError), or a user misconfiguration (Error) when this happens?
        BugError "Output scheme $scheme does not accept result files ${RESULT_FILES[*]}, use OUTPUT_URL=null if you don't want to copy them anywhere."
    fi
fi

LogPrint "Copying resulting files to $scheme location"

echo "$VERSION_INFO" >"$TMP_DIR/VERSION" || Error "Could not create $TMP_DIR/VERSION file"
RESULT_FILES+=( "$TMP_DIR/VERSION" )

local usage_readme_file=$( get_template "RESULT_usage_$OUTPUT.txt" )
if test -s $usage_readme_file ; then
    cp $v $usage_readme_file "$TMP_DIR/README" || Error "Could not copy $usage_readme_file to $TMP_DIR/README"
    RESULT_FILES+=( "$TMP_DIR/README" )
fi

# Usually RUNTIME_LOGFILE=/var/log/rear/rear-$HOSTNAME.log
# The RUNTIME_LOGFILE name is set by the main script from LOGFILE in default.conf
# but later user config files are sourced in the main script where LOGFILE can be set different
# so that the user config LOGFILE basename is used as final logfile name:
local final_logfile_name=$( basename $LOGFILE )
cat "$RUNTIME_LOGFILE" > "$TMP_DIR/$final_logfile_name" || Error "Could not copy $RUNTIME_LOGFILE to $TMP_DIR/$final_logfile_name"
RESULT_FILES+=( "$TMP_DIR/$final_logfile_name" )
LogPrint "Saving $RUNTIME_LOGFILE as $final_logfile_name to $scheme location"

# The real work (actually copying resulting files to the output location):
if scheme_supports_filesystem $scheme ; then
    # We can access the destination as a mounted filesystem. Do nothing special,
    # simply copy the output files there. (Covers stuff like nfs|cifs|usb|file|sshfs|ftpfs|davfs.)
    # This won't work for iso:// , but iso can't be a OUTPUT_URL scheme, this is checked in
    # prep/default/040_check_backup_and_output_scheme.sh
    # This covers also unknown schemes, because mount_url() will attempt to mount them and fail if this is not possible,
    # so if we got here, the URL had been mounted successfully.
    local opath
    opath=$( output_path $scheme $path )
    LogPrint "Copying result files '${RESULT_FILES[*]}' to $opath at $scheme location"
    # Copy each result file one by one to avoid usually false error exits as in
    # https://github.com/rear/rear/issues/1711#issuecomment-380009044
    # where in case of an improper RESULT_FILES array member 'cp' can error out with something like
    #   cp: will not overwrite just-created '/tmp/rear.XXX/outputfs/f121/rear-f121.log' with '/tmp/rear.XXX/tmp/rear-f121.log'
    # See
    # https://stackoverflow.com/questions/4669420/have-you-ever-got-this-message-when-moving-a-file-mv-will-not-overwrite-just-c
    # which is about the same for 'mv', how to reproduce it:
    #   mkdir a b c
    #   touch a/f b/f
    #   mv a/f b/f c/
    #     mv: will not overwrite just-created 'c/f' with 'b/f'
    # It happens because two different files with the same name would be moved to the same place with only one command.
    # The -f option won't help for this case, it only applies when there already is a target file that will be overwritten.
    # Accordingly it is sufficient (even without '-f') to copy each result file one by one:
    for result_file in "${RESULT_FILES[@]}" ; do

# note: s390 kernel copy is only through nfs
        #
        # s390 optional naming override of initrd and kernel to match the s390 filesytem naming conventions
        # on s390a there is an option to name the initrd and kernel in the form of
        # file name on s390 are in the form of name type mode
        # the name is the userid or vm name and the type is initrd or kernel
        # if the vm name (cp q userid) is HOSTA then the files written will be HOSTA kernel and HOSTA initrd
        # vars needed:
        # ZVM_NAMING      - set in local.conf, if Y then enable naming override
        # ZVM_KERNEL_NAME - keeps track of kernel name in results array
        # ARCH            - override only if ARCH is Linux-s390
        #
        # initrd name override is handled in 900_create_initramfs.sh
        # kernel name override is handled in 400_guess_kernel.sh
        # kernel name override is handled in 950_copy_result_files.sh

if [[ "$ZVM_NAMING" == "Y" && "$ARCH" == "Linux-s390" ]] ; then
           if [[ -z $opath ]] ; then
              Error "Output path is not set, please check OUTPUT_URL in local.conf."
           fi

if [ "$ZVM_KERNEL_NAME" == "$result_file" ] ; then
              VM_UID=$(vmcp q userid |awk '{ print $1 }')

if [[ -z $VM_UID ]] ; then
                 Error "VM UID is not set, VM UID is set from call to vmcp.  Please make sure vmcp is available and 'vmcp q userid' returns VM ID"
              fi

LogPrint "s390 kernel naming override: $result_file will be written as $VM_UID.kernel"
              cp $v "$result_file" $opath/$VM_UID.kernel || Error "Could not copy result file $result_file to $opath/$VM_UID.kernel at $scheme location"
           else
              cp $v "$result_file" $opath/ || Error "Could not copy result file $result_file to $opath at $scheme location"
           fi
        else
           cp $v "$result_file" $opath/ || Error "Could not copy result file $result_file to $opath at $scheme location"
        fi
    done

return 0
fi

# Filesystem access to output destination not supported, use a scheme-specific tool (rsync, lftp)
case "$scheme" in
    (fish|ftp|ftps|hftp|http|https|sftp)
        LogPrint "Copying result files '${RESULT_FILES[*]}' to $scheme location"
        Log "lftp -c $OUTPUT_LFTP_OPTIONS; open $OUTPUT_URL; mput ${RESULT_FILES[*]}"

# Make sure that destination directory exists, otherwise lftp would copy
        # RESULT_FILES into last available directory in the path.
        # e.g. OUTPUT_URL=sftp://<host_name>/iso/server1 and have "/iso/server1"
        # directory missing, would upload RESULT_FILES into sftp://<host_name>/iso/
        lftp -c "$OUTPUT_LFTP_OPTIONS; open $OUTPUT_URL; mkdir -fp ${path}"
        lftp -c "$OUTPUT_LFTP_OPTIONS; open $OUTPUT_URL; mput ${RESULT_FILES[*]}" || Error "lftp failed to transfer '${RESULT_FILES[*]}' to '$OUTPUT_URL' (lftp exit code: $?)"
        ;;
    (rsync)
        # If BACKUP = RSYNC output/RSYNC/default/900_copy_result_files.sh took care of it:
        test "$BACKUP" = "RSYNC" && return 0
        LogPrint "Copying result files '${RESULT_FILES[*]}' to $scheme location"
        Log "rsync -a $v ${RESULT_FILES[*]} ${host}:${path}"
        rsync -a $v "${RESULT_FILES[@]}" "${host}:${path}" || Error "Problem transferring result files to $OUTPUT_URL"
        ;;
    (*)
        # Should be unreachable, if we got here, it is a bug.
        # Unknown schemes are handled in mount_url(), which tries to mount them and aborts if they are unsupported.
        # If they can be mounted, they fall under the scheme_supports_filesystem branch above.
        BugError "Invalid scheme '$scheme' in '$OUTPUT_URL'."
        ;;
esac

MR SKB