From: Eric Blake <eblake@redhat.com>
To: qemu-devel@nongnu.org
Cc: Nir Soffer <nsoffer@redhat.com>,
"open list:Network Block Dev..." <qemu-block@nongnu.org>,
Vladimir Sementsov-Ogievskiy <vsementsov@virtuozzo.com>,
qemu-stable@nongnu.org, Nir Soffer <nirsof@gmail.com>
Subject: [PULL 01/20] qemu-nbd: Change default cache mode to writeback
Date: Mon, 27 Sep 2021 16:55:26 -0500 [thread overview]
Message-ID: <20210927215545.3930309-2-eblake@redhat.com> (raw)
In-Reply-To: <20210927215545.3930309-1-eblake@redhat.com>
From: Nir Soffer <nirsof@gmail.com>
Both qemu and qemu-img use writeback cache mode by default, which is
already documented in qemu(1). qemu-nbd uses writethrough cache mode by
default, and the default cache mode is not documented.
According to the qemu-nbd(8):
--cache=CACHE
The cache mode to be used with the file. See the
documentation of the emulator's -drive cache=... option for
allowed values.
qemu(1) says:
The default mode is cache=writeback.
So users have no reason to assume that qemu-nbd is using writethough
cache mode. The only hint is the painfully slow writing when using the
defaults.
Looking in git history, it seems that qemu used writethrough in the past
to support broken guests that did not flush data properly, or could not
flush due to limitations in qemu. But qemu-nbd clients can use
NBD_CMD_FLUSH to flush data, so using writethrough does not help anyone.
Change the default cache mode to writback, and document the default and
available values properly in the online help and manual.
With this change converting image via qemu-nbd is 3.5 times faster.
$ qemu-img create dst.img 50g
$ qemu-nbd -t -f raw -k /tmp/nbd.sock dst.img
Before this change:
$ hyperfine -r3 "./qemu-img convert -p -f raw -O raw -T none -W fedora34.img nbd+unix:///?socket=/tmp/nbd.sock"
Benchmark #1: ./qemu-img convert -p -f raw -O raw -T none -W fedora34.img nbd+unix:///?socket=/tmp/nbd.sock
Time (mean ± σ): 83.639 s ± 5.970 s [User: 2.733 s, System: 6.112 s]
Range (min … max): 76.749 s … 87.245 s 3 runs
After this change:
$ hyperfine -r3 "./qemu-img convert -p -f raw -O raw -T none -W fedora34.img nbd+unix:///?socket=/tmp/nbd.sock"
Benchmark #1: ./qemu-img convert -p -f raw -O raw -T none -W fedora34.img nbd+unix:///?socket=/tmp/nbd.sock
Time (mean ± σ): 23.522 s ± 0.433 s [User: 2.083 s, System: 5.475 s]
Range (min … max): 23.234 s … 24.019 s 3 runs
Users can avoid the issue by using --cache=writeback[1] but the defaults
should give good performance for the common use case.
[1] https://bugzilla.redhat.com/1990656
Signed-off-by: Nir Soffer <nsoffer@redhat.com>
Message-Id: <20210813205519.50518-1-nsoffer@redhat.com>
Reviewed-by: Eric Blake <eblake@redhat.com>
CC: qemu-stable@nongnu.org
Signed-off-by: Eric Blake <eblake@redhat.com>
---
docs/tools/qemu-nbd.rst | 6 ++++--
qemu-nbd.c | 6 ++++--
2 files changed, 8 insertions(+), 4 deletions(-)
diff --git a/docs/tools/qemu-nbd.rst b/docs/tools/qemu-nbd.rst
index e39a9f4b1a67..56e54cd44114 100644
--- a/docs/tools/qemu-nbd.rst
+++ b/docs/tools/qemu-nbd.rst
@@ -99,8 +99,10 @@ driver options if ``--image-opts`` is specified.
.. option:: --cache=CACHE
- The cache mode to be used with the file. See the documentation of
- the emulator's ``-drive cache=...`` option for allowed values.
+ The cache mode to be used with the file. Valid values are:
+ ``none``, ``writeback`` (the default), ``writethrough``,
+ ``directsync`` and ``unsafe``. See the documentation of
+ the emulator's ``-drive cache=...`` option for more info.
.. option:: -n, --nocache
diff --git a/qemu-nbd.c b/qemu-nbd.c
index 65ebec598f88..9d895ba24b1e 100644
--- a/qemu-nbd.c
+++ b/qemu-nbd.c
@@ -135,7 +135,9 @@ static void usage(const char *name)
" 'snapshot.id=[ID],snapshot.name=[NAME]', or\n"
" '[ID_OR_NAME]'\n"
" -n, --nocache disable host cache\n"
-" --cache=MODE set cache mode (none, writeback, ...)\n"
+" --cache=MODE set cache mode used to access the disk image, the\n"
+" valid options are: 'none', 'writeback' (default),\n"
+" 'writethrough', 'directsync' and 'unsafe'\n"
" --aio=MODE set AIO mode (native, io_uring or threads)\n"
" --discard=MODE set discard mode (ignore, unmap)\n"
" --detect-zeroes=MODE set detect-zeroes mode (off, on, unmap)\n"
@@ -552,7 +554,7 @@ int main(int argc, char **argv)
bool alloc_depth = false;
const char *tlscredsid = NULL;
bool imageOpts = false;
- bool writethrough = true;
+ bool writethrough = false; /* Client will flush as needed. */
bool fork_process = false;
bool list = false;
int old_stderr = -1;
--
2.31.1
next prev parent reply other threads:[~2021-09-27 21:57 UTC|newest]
Thread overview: 35+ messages / expand[flat|nested] mbox.gz Atom feed top
2021-09-27 21:55 [PULL 00/20] NBD patches through 2021-09-27 Eric Blake
2021-09-27 21:55 ` Eric Blake [this message]
2021-09-27 21:55 ` [PULL 02/20] block/io: bring request check to bdrv_co_(read, write)v_vmstate Eric Blake
2021-09-27 21:55 ` [PULL 03/20] qcow2: check request on vmstate save/load path Eric Blake
2021-09-27 21:55 ` [PULL 04/20] block: use int64_t instead of uint64_t in driver read handlers Eric Blake
2021-09-27 21:55 ` [PULL 05/20] block: use int64_t instead of uint64_t in driver write handlers Eric Blake
2021-09-27 21:55 ` [PULL 06/20] block: use int64_t instead of uint64_t in copy_range driver handlers Eric Blake
2021-09-27 21:55 ` [PULL 07/20] block: make BlockLimits::max_pwrite_zeroes 64bit Eric Blake
2021-09-27 21:55 ` [PULL 08/20] block: use int64_t instead of int in driver write_zeroes handlers Eric Blake
2021-09-27 21:55 ` [PULL 09/20] block/io: allow 64bit write-zeroes requests Eric Blake
2021-09-27 21:55 ` [PULL 10/20] block: make BlockLimits::max_pdiscard 64bit Eric Blake
2021-09-27 21:55 ` [PULL 11/20] block: use int64_t instead of int in driver discard handlers Eric Blake
2021-09-27 21:55 ` [PULL 12/20] block/io: allow 64bit discard requests Eric Blake
2021-09-27 21:55 ` [PULL 13/20] nbd/server: Allow LIST_META_CONTEXT without STRUCTURED_REPLY Eric Blake
2021-09-27 21:55 ` [PULL 14/20] nbd/client-connection: nbd_co_establish_connection(): fix non set errp Eric Blake
2021-09-27 21:55 ` [PULL 15/20] block/nbd: nbd_channel_error() shutdown channel unconditionally Eric Blake
2021-09-27 21:55 ` [PULL 16/20] block/nbd: move nbd_recv_coroutines_wake_all() up Eric Blake
2021-09-27 21:55 ` [PULL 17/20] block/nbd: refactor nbd_recv_coroutines_wake_all() Eric Blake
2021-09-27 21:55 ` [PULL 18/20] block/nbd: drop connection_co Eric Blake
2022-02-02 11:49 ` Fabian Ebner
2022-02-02 13:53 ` Eric Blake
2022-02-02 14:21 ` Hanna Reitz
2022-02-03 8:49 ` Fabian Ebner
2021-09-27 21:55 ` [PULL 19/20] block/nbd: check that received handle is valid Eric Blake
2021-09-27 21:55 ` [PULL 20/20] nbd/server: Add --selinux-label option Eric Blake
2021-09-29 8:59 ` [PULL 00/20] NBD patches through 2021-09-27 Peter Maydell
2021-09-29 12:40 ` Paolo Bonzini
2021-09-29 13:58 ` Richard Henderson
2021-09-29 15:03 ` Paolo Bonzini
2021-09-29 18:29 ` Eric Blake
2021-09-29 19:14 ` Richard W.M. Jones
2021-09-30 8:29 ` Daniel P. Berrangé
2021-09-30 8:45 ` Richard W.M. Jones
2021-09-30 14:27 ` Richard Henderson
2021-09-30 14:37 ` Richard W.M. Jones
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=20210927215545.3930309-2-eblake@redhat.com \
--to=eblake@redhat.com \
--cc=nirsof@gmail.com \
--cc=nsoffer@redhat.com \
--cc=qemu-block@nongnu.org \
--cc=qemu-devel@nongnu.org \
--cc=qemu-stable@nongnu.org \
--cc=vsementsov@virtuozzo.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 a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for NNTP newsgroup(s).