From: "Peter Kjellerstedt" <peter.kjellerstedt@axis.com>
To: "hkleynhans@fb.com" <hkleynhans@fb.com>,
"poky@lists.yoctoproject.org" <poky@lists.yoctoproject.org>
Cc: "rmikey@fb.com" <rmikey@fb.com>
Subject: Re: [poky] [PATCH] sstate: Add ZStandard compressor support
Date: Fri, 1 Oct 2021 13:13:39 +0000 [thread overview]
Message-ID: <9e6cc87314854416861d13b5d1ec887c@axis.com> (raw)
In-Reply-To: <20211001101118.2526538-1-hkleynhans@fb.com>
> -----Original Message-----
> From: poky@lists.yoctoproject.org <poky@lists.yoctoproject.org> On Behalf
> Of Henry Kleynhans via lists.yoctoproject.org
> Sent: den 1 oktober 2021 12:11
> To: poky@lists.yoctoproject.org
> Cc: hkleynhans@fb.com; rmikey@fb.com
> Subject: [poky] [PATCH] sstate: Add ZStandard compressor support
>
> This patch adds support to optionally use the Zstandard compressor for
> ssate cache files.
>
> Zstandard compression provides a significant improvement in
> decompression speed as well as improvement in compression speed and disk
> usage over the 'tgz' format in use. Furthermore, its configurable
> compression level offers a trade-off between time spent compressing
> sstate cache files and disk space used by those files. The reduced disk
> usage also contributes to saving network traffic for those sharing their
> sstate cache with others.
>
> Zstandard should therefore be a good choice when:
> * disk space is at a premium
> * network speed / resources are limited
> * the CI server can sstate packages can be created at high compression
> * less CPU on the build server should be used for sstate decompression
>
> Signed-off-by: Henry Kleynhans <hkleynhans@fb.com>
> ---
> meta/classes/sstate.bbclass | 49 +++++++++++++++++++++++-------
> scripts/sstate-cache-management.sh | 40 ++++++++++++------------
> 2 files changed, 58 insertions(+), 31 deletions(-)
>
> diff --git a/meta/classes/sstate.bbclass b/meta/classes/sstate.bbclass
> index 92a73114bb..a73d631679 100644
> --- a/meta/classes/sstate.bbclass
> +++ b/meta/classes/sstate.bbclass
> @@ -1,17 +1,30 @@
> SSTATE_VERSION = "3"
>
> +SSTATE_USE_ZSTD ?= "0"
> +SSTATE_ZSTD_CLEVEL ?= "3"
> +SSTATE_ZSTD_NTHREADS ?= "0"
Do we really need to make this configurable? Can't we just decide
to use zstd and be done with it?
> +
> SSTATE_MANIFESTS ?= "${TMPDIR}/sstate-control"
> SSTATE_MANFILEPREFIX = "${SSTATE_MANIFESTS}/manifest-${SSTATE_MANMACH}-${PN}"
>
> -def generate_sstatefn(spec, hash, taskname, siginfo, d):
> +def generate_sstate_ext(use_zstd, d):
> + if use_zstd == "1":
> + return "tar.zst"
> + return "tgz"
> +
> +def generate_sstatefn(spec, hash, taskname, siginfo, use_zstd, d):
> if taskname is None:
> return ""
> extension = ".tgz"
> + if use_zstd == "1":
> + extension = ".tar.zst"
> # 8 chars reserved for siginfo
> limit = 254 - 8
> if siginfo:
> limit = 254
> extension = ".tgz.siginfo"
> + if use_zstd == "1":
> + extension = ".tar.zst.siginfo"
> if not hash:
> hash = "INVALID"
> fn = spec + hash + "_" + taskname + extension
> @@ -33,11 +46,12 @@ def generate_sstatefn(spec, hash, taskname, siginfo, d):
> SSTATE_PKGARCH = "${PACKAGE_ARCH}"
> SSTATE_PKGSPEC = "sstate:${PN}:${PACKAGE_ARCH}${TARGET_VENDOR}-${TARGET_OS}:${PV}:${PR}:${SSTATE_PKGARCH}:${SSTATE_VERSION}:"
> SSTATE_SWSPEC = "sstate:${PN}::${PV}:${PR}::${SSTATE_VERSION}:"
> -SSTATE_PKGNAME = "${SSTATE_EXTRAPATH}${@generate_sstatefn(d.getVar('SSTATE_PKGSPEC'), d.getVar('BB_UNIHASH'), d.getVar('SSTATE_CURRTASK'), False, d)}"
> +SSTATE_PKGNAME = "${SSTATE_EXTRAPATH}${@generate_sstatefn(d.getVar('SSTATE_PKGSPEC'), d.getVar('BB_UNIHASH'), d.getVar('SSTATE_CURRTASK'), False, d.getVar('SSTATE_USE_ZSTD'), d)}"
> SSTATE_PKG = "${SSTATE_DIR}/${SSTATE_PKGNAME}"
> SSTATE_EXTRAPATH = ""
> SSTATE_EXTRAPATHWILDCARD = ""
> -SSTATE_PATHSPEC = "${SSTATE_DIR}/${SSTATE_EXTRAPATHWILDCARD}*/*/${SSTATE_PKGSPEC}*_${SSTATE_PATH_CURRTASK}.tgz*"
> +SSTATE_PKG_EXT = "${@generate_sstate_ext(d.getVar('SSTATE_USE_ZSTD'), d)}"
> +SSTATE_PATHSPEC = "${SSTATE_DIR}/${SSTATE_EXTRAPATHWILDCARD}*/*/${SSTATE_PKGSPEC}*_${SSTATE_PATH_CURRTASK}.${SSTATE_PKG_EXT}*"
>
> # explicitly make PV to depend on evaluated value of PV variable
> PV[vardepvalue] = "${PV}"
> @@ -825,12 +839,20 @@ sstate_create_package () {
> mkdir --mode=0775 -p `dirname ${SSTATE_PKG}`
> TFILE=`mktemp ${SSTATE_PKG}.XXXXXXXX`
>
> - # Use pigz if available
> - OPT="-czS"
> - if [ -x "$(command -v pigz)" ]; then
> - OPT="-I pigz -cS"
> + if [ x"${SSTATE_USE_ZSTD}" != x"0" ]; then
> + export ZSTD_CLEVEL="${SSTATE_ZSTD_CLEVEL}"
> + export ZSTD_NBTHREADS="${SSTATE_ZSTD_NTHREADS}"
> + OPT="-I zstd -cS"
> + else
> + # Use pigz if available
> + OPT="-czS"
> + if [ -x "$(command -v pigz)" ]; then
> + OPT="-I pigz -cS"
> + fi
> fi
>
> + echo "OPTS=${OPTS}"
> +
Remove debug output again.
> # Need to handle empty directories
> if [ "$(ls -A)" ]; then
> set +e
> @@ -880,7 +902,12 @@ python sstate_report_unihash() {
> # Will be run from within SSTATE_INSTDIR.
> #
> sstate_unpack_package () {
> - tar -xvzf ${SSTATE_PKG}
> + if [[ "${SSTATE_PKG}" == *.tar.zst ]]; then
Don't use [[ ]]. It is a bashism. Instead you can use:
case ${SSTATE_PKG} in
*.tar.zst)
export ZSTD_NBTHREADS="${SSTATE_ZSTD_NTHREADS}"
tar -I zstd -xvf ${SSTATE_PKG}
;;
*)
tar -xvzf ${SSTATE_PKG}
;;
esac
However, tar should be able to figure out the decompressor itself,
so it should be possible to just do:
export ZSTD_NBTHREADS="${SSTATE_ZSTD_NTHREADS}"
tar -xvf ${SSTATE_PKG}
> + export ZSTD_NBTHREADS="${SSTATE_ZSTD_NTHREADS}"
> + tar -I zstd -xvf ${SSTATE_PKG}
> + else
> + tar -xvzf ${SSTATE_PKG}
> + fi
> # update .siginfo atime on local/NFS mirror
> [ -O ${SSTATE_PKG}.siginfo ] && [ -w ${SSTATE_PKG}.siginfo ] && [ -h ${SSTATE_PKG}.siginfo ] && touch -a ${SSTATE_PKG}.siginfo
> # Use "! -w ||" to return true for read only files
> @@ -922,7 +949,7 @@ def sstate_checkhashes(sq_data, d, siginfo=False, currentcount=0, summary=True,
>
> spec, extrapath, tname = getpathcomponents(tid, d)
>
> - sstatefile = d.expand("${SSTATE_DIR}/" + extrapath + generate_sstatefn(spec, gethash(tid), tname, siginfo, d))
> + sstatefile = d.expand("${SSTATE_DIR}/" + extrapath + generate_sstatefn(spec, gethash(tid), tname, siginfo, d.getVar('SSTATE_USE_ZSTD'), d))
>
> if os.path.exists(sstatefile):
> bb.debug(2, "SState: Found valid sstate file %s" % sstatefile)
> @@ -1016,11 +1043,11 @@ def sstate_checkhashes(sq_data, d, siginfo=False, currentcount=0, summary=True,
> evdata = {'missed': [], 'found': []};
> for tid in missed:
> spec, extrapath, tname = getpathcomponents(tid, d)
> - sstatefile = d.expand(extrapath + generate_sstatefn(spec, gethash(tid), tname, False, d))
> + sstatefile = d.expand(extrapath + generate_sstatefn(spec, gethash(tid), tname, siginfo, False, d))
> evdata['missed'].append((bb.runqueue.fn_from_tid(tid), bb.runqueue.taskname_from_tid(tid), gethash(tid), sstatefile ) )
> for tid in found:
> spec, extrapath, tname = getpathcomponents(tid, d)
> - sstatefile = d.expand(extrapath + generate_sstatefn(spec, gethash(tid), tname, False, d))
> + sstatefile = d.expand(extrapath + generate_sstatefn(spec, gethash(tid), tname, siginfo, False, d))
> evdata['found'].append((bb.runqueue.fn_from_tid(tid), bb.runqueue.taskname_from_tid(tid), gethash(tid), sstatefile ) )
> bb.event.fire(bb.event.MetadataEvent("MissedSstate", evdata), d)
>
> diff --git a/scripts/sstate-cache-management.sh b/scripts/sstate-cache-management.sh
> index f1706a2229..61c7f9f763 100755
> --- a/scripts/sstate-cache-management.sh
> +++ b/scripts/sstate-cache-management.sh
> @@ -114,7 +114,7 @@ echo_error () {
> # * Add .done/.siginfo to the remove list
> # * Add destination of symlink to the remove list
> #
> -# $1: output file, others: sstate cache file (.tgz)
> +# $1: output file, others: sstate cache file (.tgz or .tar.zstd)
> gen_rmlist (){
> local rmlist_file="$1"
> shift
> @@ -131,13 +131,13 @@ gen_rmlist (){
> dest="`readlink -e $i`"
> if [ -n "$dest" ]; then
> echo $dest >> $rmlist_file
> - # Remove the .siginfo when .tgz is removed
> + # Remove the .siginfo when .tgz or .tar.zst is removed
> if [ -f "$dest.siginfo" ]; then
> echo $dest.siginfo >> $rmlist_file
> fi
> fi
> fi
> - # Add the ".tgz.done" and ".siginfo.done" (may exist in the future)
> + # Add the ".tgz.done" or ".tar.zst.done" and ".siginfo.done" (may exist in the future)
> base_fn="${i##/*/}"
> t_fn="$base_fn.done"
> s_fn="$base_fn.siginfo.done"
> @@ -188,10 +188,10 @@ remove_duplicated () {
> total_files=`find $cache_dir -name 'sstate*' | wc -l`
> # Save all the sstate files in a file
> sstate_files_list=`mktemp` || exit 1
> - find $cache_dir -name 'sstate:*:*:*:*:*:*:*.tgz*' >$sstate_files_list
> + find $cache_dir -name 'sstate:*:*:*:*:*:*:*.tgz*' -o -iname 'sstate:*:*:*:*:*:*:*.tar.zst*' >$sstate_files_list
>
> echo "Figuring out the suffixes in the sstate cache dir ... "
> - sstate_suffixes="`sed 's%.*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^_]*_\([^:]*\)\.tgz.*%\1%g' $sstate_files_list | sort -u`"
> + sstate_suffixes="`sed 's%.*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^_]*_\([^:]*\)\.\(tgz\|tar\.\zst\).*%\1%g' $sstate_files_list | sort -u`"
> echo "Done"
> echo "The following suffixes have been found in the cache dir:"
> echo $sstate_suffixes
> @@ -200,10 +200,10 @@ remove_duplicated () {
> # Using this SSTATE_PKGSPEC definition it's 6th colon separated field
> # SSTATE_PKGSPEC = "sstate:${PN}:${PACKAGE_ARCH}${TARGET_VENDOR}-${TARGET_OS}:${PV}:${PR}:${SSTATE_PKGARCH}:${SSTATE_VERSION}:"
> for arch in $all_archs; do
> - grep -q ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:$arch:[^:]*:[^:]*\.tgz$" $sstate_files_list
> + grep -q ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:$arch:[^:]*:[^:]*\.\(tgz\|tar\.\zst\)$" $sstate_files_list
> [ $? -eq 0 ] && ava_archs="$ava_archs $arch"
> # ${builder_arch}_$arch used by toolchain sstate
> - grep -q ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:${builder_arch}_$arch:[^:]*:[^:]*\.tgz$ " $sstate_files_list
> + grep -q ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:${builder_arch}_$arch:[^:]*:[^:]*\.\(tgz\|tar\.zst\)$" $sstate_files_list
> [ $? -eq 0 ] && ava_archs="$ava_archs ${builder_arch}_$arch"
> done
> echo "Done"
> @@ -219,13 +219,13 @@ remove_duplicated () {
> continue
> fi
> # Total number of files including .siginfo and .done files
> - total_files_suffix=`grep ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:_]*_$suffix\.tgz.*" $sstate_files_list | wc -l 2>/dev/null`
> - total_tgz_suffix=`grep ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:_]*_$suffix\.tgz$" $sstate_files_list | wc -l 2>/dev/null`
> + total_files_suffix=`grep ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:_]*_$suffix\.\(tgz\|tar\.zst\).*" $sstate_files_list | wc -l 2>/dev/null`
> + total_archive_suffix=`grep ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:_]*_$suffix\.\(tgz\|tar\.zst\)$" $sstate_files_list | wc -l 2>/dev/null`
> # Save the file list to a file, some suffix's file may not exist
> - grep ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:_]*_$suffix\.tgz.*" $sstate_files_list >$list_suffix 2>/dev/null
> - local deleted_tgz=0
> + grep ".*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:_]*_$suffix\.\(tgz\|tar\.zst\).*" $sstate_files_list >$list_suffix 2>/dev/null
> + local deleted_archives=0
> local deleted_files=0
> - for ext in tgz tgz.siginfo tgz.done; do
> + for ext in tgz tgz.siginfo tgz.done tar.zst tar.zst.siginfo tar.zst.done; do
> echo "Figuring out the sstate:xxx_$suffix.$ext ... "
> # Uniq BPNs
> file_names=`for arch in $ava_archs ""; do
> @@ -268,19 +268,19 @@ remove_duplicated () {
> done
> done
> done
> - deleted_tgz=`cat $rm_list.* 2>/dev/null | grep ".tgz$" | wc -l`
> + deleted_archives=`cat $rm_list.* 2>/dev/null | grep ".\(tgz\|tar\.zst\)$" | wc -l`
> deleted_files=`cat $rm_list.* 2>/dev/null | wc -l`
> [ "$deleted_files" -gt 0 -a $debug -gt 0 ] && cat $rm_list.*
> - echo "($deleted_tgz out of $total_tgz_suffix .tgz files for $suffix suffix will be removed or $deleted_files out of $total_files_suffix when counting also .siginfo and .done files)"
> + echo "($deleted_archives out of $total_archives_suffix .tgz or .tar.zst files for $suffix suffix will be removed or $deleted_files out of $total_files_suffix when counting also .siginfo and .done files)"
> let total_deleted=$total_deleted+$deleted_files
> done
> - deleted_tgz=0
> + deleted_archives=0
> rm_old_list=$remove_listdir/sstate-old-filenames
> - find $cache_dir -name 'sstate-*.tgz' >$rm_old_list
> - [ -s "$rm_old_list" ] && deleted_tgz=`cat $rm_old_list | grep ".tgz$" | wc -l`
> + find $cache_dir -name 'sstate-*.tgz' -o -name 'sstate-*.tar.zst' >$rm_old_list
> + [ -s "$rm_old_list" ] && deleted_archives=`cat $rm_old_list | grep ".\(tgz\|tar\.zst\)$" | wc -l`
> [ -s "$rm_old_list" ] && deleted_files=`cat $rm_old_list | wc -l`
> [ -s "$rm_old_list" -a $debug -gt 0 ] && cat $rm_old_list
> - echo "($deleted_tgz .tgz files with old sstate-* filenames will be removed or $deleted_files when counting also .siginfo and .done files)"
> + echo "($deleted_archives .tgz or .tar.zst files with old sstate-* filenames will be removed or $deleted_files when counting also .siginfo and .done files)"
> let total_deleted=$total_deleted+$deleted_files
>
> rm -f $list_suffix
> @@ -289,7 +289,7 @@ remove_duplicated () {
> read_confirm
> if [ "$confirm" = "y" -o "$confirm" = "Y" ]; then
> for list in `ls $remove_listdir/`; do
> - echo "Removing $list.tgz (`cat $remove_listdir/$list | wc -w` files) ... "
> + echo "Removing $list archive (`cat $remove_listdir/$list | wc -w` files) ... "
> # Remove them one by one to avoid the argument list too long error
> for i in `cat $remove_listdir/$list`; do
> rm -f $verbose $i
> @@ -322,7 +322,7 @@ rm_by_stamps (){
> find $cache_dir -type f -name 'sstate*' | sort -u -o $cache_list
>
> echo "Figuring out the suffixes in the sstate cache dir ... "
> - local sstate_suffixes="`sed 's%.*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^_]*_\([^:]*\)\.tgz.*%\1%g' $cache_list | sort -u`"
> + local sstate_suffixes="`sed 's%.*/sstate:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^:]*:[^_]*_\([^:]*\)\.\(tgz\|tar\.zst\).*%\1%g' $cache_list | sort -u`"
> echo "Done"
> echo "The following suffixes have been found in the cache dir:"
> echo $sstate_suffixes
> --
> 2.30.2
//Peter
next prev parent reply other threads:[~2021-10-01 13:13 UTC|newest]
Thread overview: 6+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-10-01 10:11 [PATCH] sstate: Add ZStandard compressor support hkleynhans
2021-10-01 13:13 ` Peter Kjellerstedt [this message]
2021-10-01 13:16 ` [poky] " Richard Purdie
2021-10-01 13:44 ` Joshua Watt
2021-10-01 14:19 ` Henry Kleynhans
2021-10-01 13:52 ` Joshua Watt
Reply instructions:
You may reply publicly to this message via plain-text email
using any one of the following methods:
* Save the following mbox file, import it into your mail client,
and reply-to-all from there: mbox
Avoid top-posting and favor interleaved quoting:
https://en.wikipedia.org/wiki/Posting_style#Interleaved_style
* Reply using the --to, --cc, and --in-reply-to
switches of git-send-email(1):
git send-email \
--in-reply-to=9e6cc87314854416861d13b5d1ec887c@axis.com \
--to=peter.kjellerstedt@axis.com \
--cc=hkleynhans@fb.com \
--cc=poky@lists.yoctoproject.org \
--cc=rmikey@fb.com \
/path/to/YOUR_REPLY
https://kernel.org/pub/software/scm/git/docs/git-send-email.html
* If your mail client supports setting the In-Reply-To header
via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line
before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.