将 GNU 并行命令与 gfind 结合使用以获得 gupdatedb 工具的运行时间

Using GNU parallel command with gfind to gain in runtime for gupdatedb tool

我跟前面的一样post

我想构建 gupdatedb 数据库,包含来自主根 / 的所有内容,但下面列出的 PRUNEPATHS 除外。我正在使用 MacOS 10.15 Catalina。

所以,我尝试修改 MacOS 10.15 上的 gupdatedb 脚本以从这样的 parallel 命令中受益(注意 # : A2 部分):

# : A2
cat | parallel -j32 $find {} $SEARCHPATHS $FINDOPTIONS \
    \( $prunefs_exp -type d -regex "$PRUNEREGEX" \) \
    -prune -o $print_option * :::

如果我不使用 cat |,我会收到以下警告消息:

parallel: Warning: Input is read from the terminal. You are either an expert
parallel: Warning: (in which case: YOU ARE AWESOME!) or maybe you forgot
parallel: Warning: ::: or :::: or -a or to pipe data into parallel. If so
parallel: Warning: consider going through the tutorial: man parallel_tutorial
parallel: Warning: Press CTRL-D to exit.

并且进程似乎挂起。

不幸的是,$find = gfind 的多个线程似乎无法同时 运行 :

我已经启动了这样的脚本:sudo time gupdatedb

及以下结果:ps aux | grep find :

root             84865   0.0  0.0  4459044  15828 s002  S+    1:43PM   0:00.10 perl /usr/local/bin/parallel -j32 /usr/local/Cellar/findutils/4.7.0/bin/gfind {} / ( -fstype 9P -o -fstype NFS -o -fstype afs -o -fstype autofs -o -fstype cifs -o -fstype coda -o -fstype devfs -o -fstype devpts -o -fstype ftpfs -o -fstype iso9660 -o -fstype mfs -o -fstype ncpfs -o -fstype nfs -o -fstype nfs4 -o -fstype proc -o -fstype shfs -o -fstype smbfs -o -fstype sysfs -o -type d -regex \(^/afs$\)\|\(^/amd$\)\|\(^/proc$\)\|\(^/sfs$\)\|\(^/tmp$\)\|\(^/usr/tmp$\)\|\(^/var/tmp$\)\|\(^/Volumes$\) ) -prune -o -print0 Applications Library System Users Volumes bin cores dev etc home opt private sbin tmp usr var :::
root             84863   0.0  0.0  4268280    796 s002  S+    1:43PM   0:00.00 /usr/local/Cellar/findutils/4.7.0/libexec/gfrcode -0
root             84861   0.0  0.0  4282172    708 s002  S+    1:43PM   0:00.00 /bin/sh /usr/local/Cellar/findutils/4.7.0/libexec/bin/gupdatedb
root             84853   0.0  0.0  4273980   1164 s002  S+    1:43PM   0:00.01 /bin/sh /usr/local/Cellar/findutils/4.7.0/libexec/bin/gupdatedb
root             84850   0.0  0.0  5396228  10288 s008  S+    1:43PM   0:00.27 vim /usr/local/Cellar/findutils/4.7.0/libexec/bin/gupdatedb
root             84849   0.0  0.0  4788896   6740 s008  S+    1:43PM   0:00.03 sudo vim /usr/local/Cellar/findutils/4.7.0/libexec/bin/gupdatedb

最后,数据库可能无法构建,我正在检查 /usr/local/var/locate/locatedb.n/usr/local/var/locate/locatedb 的大小,但没有任何变化。

我在 parallel 中使用的语法有什么问题? (特别是,我不知道如何处理命令的 ... ::: options 部分)

PS : 我设置在 gupdatedb :

# Directories to not put in the database, which would otherwise be.
: ${PRUNEPATHS="
/afs
/amd
/proc
/sfs
/tmp
/usr/tmp
/var/tmp
/Volumes
"}

# You can set these in the environment, or use command-line options,
# to override their defaults:

# Any global options for find?
: ${FINDOPTIONS=}

# What shell shoud we use?  We should use a POSIX-ish sh.
: ${SHELL="/bin/sh"}

# Non-network directories to put in the database.
: ${SEARCHPATHS="/"}

更新 1

更准确地说,这里是 post 我要求与 parallel/find 夫妇进行潜在优化(并行化)的地方:

我想做同样的优化,但针对脚本 gupdatedb

更新 2

我听从了 :

的建议

进入 gupdatedb 关于我的问题的默认命令是:

$find $SEARCHPATHS $FINDOPTIONS \
 \( $prunefs_exp \
 -type d -regex "$PRUNEREGEX" \) -prune -o $print_option

所以,我刚刚修改成这样:

parallel -j32 $find {} $SEARCHPATHS $FINDOPTIONS \
    \( $prunefs_exp \
    -type d -regex "$PRUNEREGEX" \) -prune -o $print_option ::: /

我收到以下错误:

/bin/sh: -c: line 0: syntax error near unexpected token `('
/bin/sh: -c: line 0: `/usr/local/Cellar/findutils/4.7.0/bin/gfind / / ( -fstype 9P -o -fstype NFS -o -fstype afs -o -fstype autofs -o -fstype cifs -o -fstype coda -o -fstype devfs -o -fstype devpts -o -fstype ftpfs -o -fstype iso9660 -o -fstype mfs -o -fstype ncpfs -o -fstype nfs -o -fstype nfs4 -o -fstype proc -o -fstype shfs -o -fstype smbfs -o -fstype sysfs -o -type d -regex \(^/private/tmp$\)\|\(^/private/var/folders$\)\|\(^/private/var/tmp$\)\|\(^*/Backups.backupdb$\)\|\(^/System$\)\|\(^/Volumes$\) ) -prune -o -print0'

这里可能有什么问题?

更新 3

这里是脚本 gupdatedb,您可以从第 300 行看到我的不同尝试:

#! /bin/sh
# updatedb -- build a locate pathname database
# Copyright (C) 1994-2019 Free Software Foundation, Inc.
#
# This program 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.
#
# This program 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 this program.  If not, see <https://www.gnu.org/licenses/>.

# csh original by James Woods; sh conversion by David MacKenzie.

#exec 2> /tmp/updatedb-trace.txt
#set -x

version='
updatedb (GNU findutils) 4.7.0
Copyright (C) 1994-2019 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <https://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.

Written by Eric B. Decker, James Youngman, and Kevin Dalley.
'

# File path names are not actually text, anyway (since there is no
# mechanism to enforce any constraint that the basename of a
# subdirectory has the same character encoding as the basename of its
# parent).  The practical effect is that, depending on the way a
# particular system is configured and the content of its filesystem,
# passing all the file names in the system through "sort" may generate
# character encoding errors in text-based tools like "sort".  To avoid
# this, we set LC_ALL=C.  This will, presumably, not work perfectly on
# systems where LC_ALL is not the way to do locale configuration or
# some other seting can override this.
LC_ALL=C
export LC_ALL

# We can't use substitution on PACKAGE_URL below because it
# (correctly) points to https://www.gnu.org/software/findutils/ instead
# of the bug reporting page.
usage="\
Usage: [=18=] [--findoptions='-option1 -option2...']
       [--localpaths='dir1 dir2...'] [--netpaths='dir1 dir2...']
       [--prunepaths='dir1 dir2...'] [--prunefs='fs1 fs2...']
       [--output=dbfile] [--netuser=user] [--localuser=user]
       [--dbformat] [--version] [--help]

Please see also the documentation at http://www.gnu.org/software/findutils/.
Report (and track progress on fixing) bugs in the updatedb
program via the GNU findutils bug-reporting page at
https://savannah.gnu.org/bugs/?group=findutils or, if
you have no web access, by sending email to <bug-findutils@gnu.org>.
"
changeto=/

for arg
do
  # If we are unable to fork, the back-tick operator will
  # fail (and the shell will emit an error message).  When
  # this happens, we exit with error value 71 (EX_OSERR).
  # Alternative candidate - 75, EX_TEMPFAIL.
  opt=`echo $arg|sed 's/^\([^=]*\).*//'`  || exit 71
  val=`echo $arg|sed 's/^[^=]*=\(.*\)//'` || exit 71
  case "$opt" in
    --findoptions) FINDOPTIONS="$val" ;;
    --localpaths) SEARCHPATHS="$val" ;;
    --netpaths) NETPATHS="$val" ;;
    --prunepaths) PRUNEPATHS="$val" ;;
    --prunefs) PRUNEFS="$val" ;;
    --output) LOCATE_DB="$val" ;;
    --netuser) NETUSER="$val" ;;
    --localuser) LOCALUSER="$val" ;;
    --changecwd)  changeto="$val" ;;
    --dbformat)   dbformat="$val" ;;
    --version) fail=0; echo "$version" || fail=1; exit $fail ;;
    --help)    fail=0; echo "$usage"   || fail=1; exit $fail ;;
    *) echo "updatedb: invalid option $opt
Try '[=18=] --help' for more information." >&2
       exit 1 ;;
  esac
done

frcode_options=""
case "$dbformat" in
    "")
        # Default, use LOCATE02
        ;;
    LOCATE02)
        ;;
    slocate)
        frcode_options="$frcode_options -S 1"
        ;;
    *)
        # The "old" database format is no longer supported.
        echo "Unsupported locate database format ${dbformat}: Supported formats are:" >&2
        echo "LOCATE02, slocate" >&2
        exit 1
esac


if true
then
    sort="/usr/bin/sort -z"
    print_option="-print0"
    frcode_options="$frcode_options -0"
else
    sort="/usr/bin/sort"
    print_option="-print"
fi

getuid() {
    # format of "id" output is ...
    # uid=1(daemon) gid=1(other)
    # for `id's that don't understand -u
    id | cut -d'(' -f 1 | cut -d'=' -f2
}

# figure out if su supports the -s option
select_shell() {
    if su "" -s $SHELL -c false < /dev/null  ; then
    # No.
    echo ""
    else
    if su "" -s $SHELL -c true < /dev/null  ; then
        # Yes.
        echo "-s $SHELL"
        else
        # su is unconditionally failing.  We won't be able to
        # figure out what is wrong, so be conservative.
        echo ""
    fi
    fi
}


# You can set these in the environment, or use command-line options,
# to override their defaults:

# Any global options for find?
: ${FINDOPTIONS="-mindepth 1 -maxdepth 1"}
#: ${FINDOPTIONS=""}

# What shell shoud we use?  We should use a POSIX-ish sh.
: ${SHELL="/bin/sh"}

# Non-network directories to put in the database.
: ${SEARCHPATHS="/"}

# Network (NFS, AFS, RFS, etc.) directories to put in the database.
: ${NETPATHS=}

# Directories to not put in the database, which would otherwise be.
: ${PRUNEPATHS="
/afs
/amd
/proc
/sfs
/tmp
/usr/tmp
/var/tmp
"}

# Trailing slashes result in regex items that are never matched, which
# is not what the user will expect.   Therefore we now reject such
# constructs.
for p in $PRUNEPATHS; do
    case "$p" in
    /*/)   echo "[=18=]: $p: pruned paths should not contain trailing slashes" >&2
           exit 1
    esac
done

# The same, in the form of a regex that find can use.
test -z "$PRUNEREGEX" &&
  PRUNEREGEX=`echo $PRUNEPATHS|sed -e 's,^,\\(^,' -e 's, ,$\\)\\|\\(^,g' -e 's,$,$\\),'`

# The database file to build.
: ${LOCATE_DB=/usr/local/var/locate/locatedb}

# Directory to hold intermediate files.
if test -z "$TMPDIR"; then
  if test -d /var/tmp; then
    : ${TMPDIR=/var/tmp}
  elif test -d /usr/tmp; then
    : ${TMPDIR=/usr/tmp}
  else
    : ${TMPDIR=/tmp}
  fi
fi
export TMPDIR

# The user to search network directories as.
: ${NETUSER=daemon}

# The directory containing the subprograms.
if test -n "$LIBEXECDIR" ; then
    : LIBEXECDIR already set, do nothing
else
    : ${LIBEXECDIR=/usr/local/Cellar/findutils/4.7.0/libexec}
fi

# The directory containing find.
if test -n "$BINDIR" ; then
    : BINDIR already set, do nothing
else
    : ${BINDIR=/usr/local/Cellar/findutils/4.7.0/bin}
fi

# The names of the utilities to run to build the database.
: ${find:=${BINDIR}/gfind}
: ${frcode:=${LIBEXECDIR}/gfrcode}

make_tempdir () {
    # This implementation is adapted from the GNU Autoconf manual.
    {
        tmp=`
    (umask 077 && mktemp -d "$TMPDIR/updatedbXXXXXX") 2>/dev/null
    ` &&
        test -n "$tmp" && test -d "$tmp"
    } || {
    # This method is less secure than mktemp -d, but it's a fallback.
    #
    # We use $$ as well as $RANDOM since $RANDOM may not be available.
    # We also add a time-dependent suffix.  This is actually somewhat
    # predictable, but then so is $$.  POSIX does not require date to
    # support +%N.
    ts=`date +%N%S || date +%S 2>/dev/null`
        tmp="$TMPDIR"/updatedb"$$"-"${RANDOM:-}${ts}"
        (umask 077 && mkdir "$tmp")
    }
    echo "$tmp"
}

checkbinary () {
    if test -x "" ; then
    : ok
    else
      eval echo "updatedb needs to be able to execute , but cannot." >&2
      exit 1
    fi
}

for binary in $find $frcode
do
  checkbinary $binary
done


: ${PRUNEFS="
9P
NFS
afs
autofs
cifs
coda
devfs
devpts
ftpfs
iso9660
mfs
ncpfs
nfs
nfs4
proc
shfs
smbfs
sysfs
"}

if test -n "$PRUNEFS"; then
prunefs_exp=`echo $PRUNEFS |sed -e 's/\([^ ][^ ]*\)/-o -fstype /g' \
 -e 's/-o //' -e 's/$/ -o/'`
else
  prunefs_exp=''
fi

# Make and code the file list.
# Sort case insensitively for users' convenience.

rm -f $LOCATE_DB.n
trap 'rm -f $LOCATE_DB.n; exit' HUP TERM

if {
cd "$changeto"
if test -n "$SEARCHPATHS"; then
  if [ "$LOCALUSER" != "" ]; then
    # : A1
    su $LOCALUSER `select_shell $LOCALUSER` -c \
    "$find $SEARCHPATHS $FINDOPTIONS \
     \( $prunefs_exp \
     -type d -regex '$PRUNEREGEX' \) -prune -o $print_option"
  else
    # : A2
    # ORIGINAL VERSION : sequential find
    #$find $SEARCHPATHS $FINDOPTIONS \
    # \( $prunefs_exp \
    # -type d -regex "$PRUNEREGEX" \) -prune -o $print_option ::: /

    # Parallel version 1
    #parallel -j 32 $find $SEARCHPATHS $FINDOPTIONS \
    # \( $prunefs_exp \
    # -type d -regex "$PRUNEREGEX" \) -prune -o $print_option ::: /
    
    # Parallel version 2
    parallel -j 32 $find {} $FINDOPTIONS \
    $prunefs_exp -type d -regex $PRUNEREGEX -prune -o $print_option ::: */*
  fi
fi

if test -n "$NETPATHS"; then
myuid=`getuid`
if [ "$myuid" = 0 ]; then
    # : A3
    su $NETUSER `select_shell $NETUSER` -c \
     "$find $NETPATHS $FINDOPTIONS \( -type d -regex '$PRUNEREGEX' -prune \) -o $print_option" ||
    exit $?
  else
    # : A4
    $find $NETPATHS $FINDOPTIONS \( -type d -regex "$PRUNEREGEX" -prune \) -o $print_option ||
    exit $?
  fi
fi
} | $sort | $frcode $frcode_options > $LOCATE_DB.n
then
    : OK so far
    true
else
    rv=$?
    echo "Failed to generate $LOCATE_DB.n" >&2
    rm -f $LOCATE_DB.n
    exit $rv
fi

# To avoid breaking locate while this script is running, put the
# results in a temp file, then rename it atomically.
if test -s $LOCATE_DB.n; then
  chmod 644 ${LOCATE_DB}.n
  mv ${LOCATE_DB}.n $LOCATE_DB
else
  echo "updatedb: new database would be empty" >&2
  rm -f $LOCATE_DB.n
fi

exit 0

我像这样启动 gupdatedb 命令:

sudo gupdatedb --prunepaths='/private/tmp /private/var/folders /private/var/tmp */Backups.backupdb /System /Volumes' --localpaths='/' --output=$HOME/locatedb_gupdatedb_PARALLEL

更新 4

我的赏金明天到期。使用默认 gupdatedb,所有索引大约需要 30 分钟。如果我能够正确使用 parallelgupdatedb 脚本的核心,即当后者使用 gfind 命令进行索引时,我可以期望哪个增益因子?

最后一个请求:如何修复错误:

/bin/sh: -c: line 0: syntax error near unexpected token `('
/bin/sh: -c: line 0: `/usr/local/Cellar/findutils/4.7.0/bin/gfind / / ( -fstype 9P -o -fstype NFS -o -fstype afs -o -fstype autofs -o -fstype cifs -o -fstype coda -o -fstype devfs -o -fstype devpts -o -fstype ftpfs -o -fstype iso9660 -o -fstype mfs -o -fstype ncpfs -o -fstype nfs -o -fstype nfs4 -o -fstype proc -o -fstype shfs -o -fstype smbfs -o -fstype sysfs -o -type d -regex \(^/private/tmp$\)\|\(^/private/var/folders$\)\|\(^/private/var/tmp$\)\|\(^*/Backups.backupdb$\)\|\(^/System$\)\|\(^/Volumes$\) ) -prune -o -print0'

使用命令:

parallel -j32 $find {} $FINDOPTIONS \
    \( $prunefs_exp \
    -type d -regex "$PRUNEREGEX" \) -prune -o $print_option ::: /

?

如果后面没有任何内容,则不需要 :::,如果没有任何来源,{} 也毫无意义。如果没有关于您究竟想要并行化什么的更多信息,我们无法真正告诉您应该使用什么。

但是,例如,如果您想要 运行 在 /etc/usr/bin 和 [=18= 中的每个 find ],看起来像

parallel find {} -options ::: /etc /usr /bin /opt

不用:::也可以这样表示:

printf '%s\n' /etc /usr /bin /opt |
parallel find {} -options

所以 ::: 的目的基本上是说“我想在命令行上指定要并行化的东西,而不是在标准输入上接收它们”;但如果您不提供此信息,无论哪种方式,parallel 都不知道用什么替换 {}

我并不是说这种特殊用途对您的用例有意义,只是希望澄清文档 ()。

要通过使用并行获得任何有意义的加速,您需要确保您有资源来加快进程。这里有两个挑战:

  1. updatedb 进程受 IO 限制。通常,您使用并行来利用 multi-core 系统,并将 CPU 绑定进程分布在多个内核上。
  2. updatedb 进程需要独占访问数据库(通常在/var/lib/mlcoate/mlocate.db 中)。即使将 updatedb 拆分到多个内核中有任何好处,您也必须将输出放入多个数据库中。此方法将需要传递所有数据库名称(用“:”分隔以使用“-d”定位)

除非您的系统有多个磁盘驱动器(或者您正在访问网络驱动器),否则您从 运行ning 并行查找中获得的收益很少。

如果您的系统有多个磁盘驱动器(and/or 网络驱动器),您可以运行 每个文件系统并行,使用像

这样的脚本

假设您在 /mnt/disk1、/mnt/disk2

上安装了 2 个额外的磁盘
  # Index root
updatedb --output=/var/lib/mlocate/local.db -E '/mnt/disk1 /mnt/disk2' &
  # Index 1st extra disk (or network drive)
updatedb --output=/var/lib/mlocate/disk1.db -U /mnt/disk1 &
  # Index 2nd extra disk (or network drive)
updatedb --output=/var/lib/mlocate/disk2.db -U /mnt/disk2 &
wait

您应该将环境变量LOCATE_PATH设置为指向所有数据库 导出

LOCATE_PATH=/var/lib/mlocate/local.db:/var/lib/mlocate/disk1.db:/var/lib/mlocate/disk2.db
locate ...