When a vdev starts getting I/O or checksum errors it is now
possible to automatically rebuild to a hot spare device.
To cleanly support this functionality in a shell script some
additional information was added to all zevent ereports which
include a vdev. This covers both io and checksum zevents but
may be used but other scripts.
In the Illumos FMA solution the same information is required
but it is retrieved through the libzfs library interface.
Specifically the following members were added:
vdev_spare_paths - List of vdev paths for all hot spares.
vdev_spare_guids - List of vdev guids for all hot spares.
vdev_read_errors - Read errors for the problematic vdev
vdev_write_errors - Write errors for the problematic vdev
vdev_cksum_errors - Checksum errors for the problematic vdev.
By default the required hot spare scripts are installed but this
functionality is disabled. To enable hot sparing uncomment the
ZED_SPARE_ON_IO_ERRORS and ZED_SPARE_ON_CHECKSUM_ERRORS in the
/etc/zfs/zed.d/zed.rc configuration file.
These scripts do no add support for the autoexpand property. At
a minimum this requires adding a new udev rule to detect when
a new device is added to the system. It also requires that the
autoexpand policy be ported from Illumos, see:
https://github.com/illumos/illumos-gate/blob/master/usr/src/cmd/syseventd/modules/zfs_mod/zfs_mod.c
Support for detecting the correct name of a vdev when it's not
a whole disk was added by Turbo Fredriksson.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Signed-off-by: Turbo Fredriksson <turbo@bayour.com>
Issue #2
This functionality has always been missing. But until now there
were no zevents which included an array of strings so it wasn't
missed. However, that's now changed so to ensure this information
is output correctly by 'zpool events -v' the DATA_TYPE_STRING_ARRAY
has been implemented.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Issue #2
Several of the zfs utilities allow you to pass a vdev's guid rather
than the device name. However, the utilities are not consistent in
how they parse that guid. For example, 'zinject' expects the guid
to be passed as a hex value while 'zpool replace' wants it as a
decimal. The user is forced to just know what format to use.
This patch improve things by making the parsing more tolerant.
When strtol(3) is called using 0 for the base, rather than say
10 or 16, it will then accept hex, decimal, or octal input based
on the prefix. From the man page.
If base is zero or 16, the string may then include a "0x"
prefix, and the number will be read in base 16; otherwise,
a zero base is taken as 10 (decimal) unless the next character
is '0', in which case it is taken as 8 (octal).
NOTE: There may be additional conversions not caught be this patch.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Issue #2
This commit adds a systemd unit file for zed.service and integrates
it into the zfs.target from commit 881f45c.
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2108
Issue #2
zed monitors ZFS events. When a zevent is posted, zed will run any
scripts that have been enabled for the corresponding zevent class.
Multiple scripts may be invoked for a given zevent. The zevent
nvpairs are passed to the scripts as environment variables.
Events are processed synchronously by the single thread, and there is
no maximum timeout for script execution. Consequently, a misbehaving
script can delay (or forever block) the processing of subsequent
zevents. Plans are to address this in future commits.
Initial scripts have been developed to log events to syslog
and send email in response to checksum/data/io errors and
resilver.finish/scrub.finish events. By default, email will only
be sent if the ZED_EMAIL variable is configured in zed.rc (which is
serving as a config file of sorts until a proper configuration file
is implemented).
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2
zpool_events_next() can be called in blocking mode by specifying a
non-zero value for the "block" parameter. However, the design of
the ZFS Event Daemon (zed) requires additional functionality from
zpool_events_next(). Instead of adding additional arguments to the
function, it makes more sense to use flags that can be bitwise-or'd
together.
This commit replaces the zpool_events_next() int "block" parameter with
an unsigned bitwise "flags" parameter. It also defines ZEVENT_NONE
to specify the default behavior. Since non-blocking mode can be
specified with the existing ZEVENT_NONBLOCK flag, the default behavior
becomes blocking mode. This, in effect, inverts the previous use
of the "block" parameter. Existing callers of zpool_events_next()
have been modified to check for the ZEVENT_NONBLOCK flag.
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2
Add macro definitions to AM_CPPFLAGS to propagate makefile installation
directory variables for libexecdir, runstatedir, sbindir, and
sysconfdir.
https://www.gnu.org/software/autoconf/manual/autoconf-2.69/html_node/Installation-Directory-Variables.html
A corollary is that you should not use these variables except
in makefiles. For instance, instead of trying to evaluate
datadir in configure and hard-coding it in makefiles using e.g.,
'AC_DEFINE_UNQUOTED([DATADIR], ["$datadir"], [Data directory.])',
you should add -DDATADIR='$(datadir)' to your makefile's definition
of CPPFLAGS (AM_CPPFLAGS if you are also using Automake).
The runstatedir directory is for "installing data files which the
programs modify while they run, that pertain to one specific machine,
and which need not persist longer than the execution of the program".
https://www.gnu.org/prep/standards/html_node/Directory-Variables.html
It will be defined by autoconf 2.70 or later, and default to
"$(localstatedir)/run".
http://git.savannah.gnu.org/gitweb/?p=autoconf.git;a=commit;h=a197431414088a417b407b9b20583b2e8f7363bd
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2
Due to the very poorly chosen argument name 'cleanup_fd' it was
completely unclear that this file descriptor is used to track the
current cursor location. When the file descriptor is created by
opening ZFS_DEV a private cursor is created in the kernel for the
returned file descriptor. Subsequent calls to zpool_events_next()
and zpool_events_seek() then require the file descriptor as an
argument to reposition the cursor. When the file descriptor is
closed the kernel state tracking the cursor is destroyed.
This patch contains no functional change, it just changes a
few variable names and clarifies the documentation.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Issue #2
The ZFS_IOC_EVENTS_SEEK ioctl was added to allow user space callers
to seek around the zevent file descriptor by EID. When a specific
EID is passed and it exists the cursor will be positioned there.
If the EID is no longer cached by the kernel ENOENT is returned.
The caller may also pass ZEVENT_SEEK_START or ZEVENT_SEEK_END to seek
to those respective locations.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Issue #2
Tagging each zevent with a unique monotonically increasing EID
(Event IDentifier) provides the required infrastructure for a user
space daemon to reliably process zevents. By writing the EID to
persistent storage the daemon can safely resume where it left off
in the event stream when it's restarted.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Chris Dunlap <cdunlap@llnl.gov>
Issue #2
Changing properties with "zfs inherit" should cause the datasets
to be remounted. This ensures that the modified property values
will be propagated in to the filesystem namespace where they can
be enforced. This change is modeled after an identical fix made
to zfs_prop_set().
Signed-off-by: Gunnar Beutner <gunnar@beutner.name>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2201
torvalds/linux@8077c0d983 added a
__must_check to the bdi_setup_and_register(), which caused our autotools
check to break. zfsonlinux/zfs@729210564a
was intended to correct that, but it depended on -Wno-unused-result,
which is unrecognized in older GCC versions. That commit has been
reverted in favor of a solution that does not require
-Wno-unused-result.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2102Closes#2135
Older GCC versions do not obey -Wno-unused-result. This reverts commit
729210564a in favor of a solution that
does not require -Wno-unused-result.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #1906
3120 zinject hangs in zfsdev_ioctl() due to uninitialized zc
Reviewed by: Richard Lowe <richlowe@richlowe.net>
Reviewed by: Eric Schrock <eric.schrock@delphix.com>
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Approved by: Richard Lowe <richlowe@richlowe.net>
References:
https://www.illumos.org/issues/3120illumos/illumos-gate@f4c46b1eda
Ported-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2152
When testing compiler flags, we only need to do compile test. Otherwise,
configure will fail with "configure: error: cannot run test program while
cross compiling" when cross compiling.
Signed-off-by: Chunwei Chen <tuxoko@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2191
Originally, users had to handle spa namespace collisions by either
exporting the already imported pool or by specifying a new name for the
pool with a conflicting name. In the case of root pools from virtual
guests, neither approach to collision resolution is reasonable. This is
addressed by extending the new name syntax with a -t option to specify
that the new name is temporary. When specified, this sets an internal
flag that is passed into the kernel to tell it that all label updates
should refer to the name used in the original label. Consequently, the
original pool name will be retained on export.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2189
Valgrind suggests that the address we are returning is not properly
aligned, so lets add an assertion.
==87740== Address 0x1012a22a is 554 bytes inside a block of size 4,096
alloc'd
==87740== at 0x4C2BBA0: memalign (in
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==87740== by 0x4C2BCC7: posix_memalign (in
/usr/lib64/valgrind/vgpreload_memcheck-amd64-linux.so)
==87740== by 0x52FA845: zio_buf_alloc (umem.h:101)
==87740== by 0x52F6226: zil_alloc_lwb (zil.c:463)
==87740== by 0x52F8559: zil_commit (zil.c:566)
==87740== by 0x40611D: ztest_freeze (ztest.c:5909)
==87740== by 0x4066A7: ztest_init (ztest.c:6048)
==87740== by 0x407AF4: main (ztest.c:6226)
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2174
Stopping arcstat.py with ^C always ends up with error:
TypeError: sighandler() takes no arguments (2 given)
Since no special signal handling was done in sighandler(),
it's simpler to just set SIGINT handler to SIG_DFL, which
terminates the script.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Isaac Huang <he.huang@intel.com>
Closes#2179
Remove the redundant call to zfs_unmount_snap() which was being
done after char array was freed,
This fixes an upstream regression that was introduced in commit
zfsonlinux/zfs@d09f25dc66, which
ported the Illumos 3744 changes.
Signed-off-by: Andrey Vesnovaty <andrey.vesnovaty@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes#2156
This is just a small bit of cleanup to ensure ztest fails early
on systems where mmap(2) is not functioning. For the automated
testing which is the primary consumer of ztest there is no
functional change because debugging is always enabled.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2177
Valgrind complained about this and it's absolutely right. The
props nvlist was not being freed in ztest_init.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes#2174
Stopping arcstat.py with ^C always ends up with error:
TypeError: sighandler() takes no arguments (2 given)
This patch corrects the error by updating the signal handler
to take the correct number of arguments.
Signed-off-by: Isaac Huang <he.huang@intel.com>
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2182
Today arcstat.py prints one header every hdr_intr (20 by default)
lines. It would be more consistent with out utilities like vmstat
if hdr_intr defaulted to terminal window size, i.e. one header
per screenful of outputs.
Signed-off-by: Isaac Huang <he.huang@intel.com>
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2183
In Debian GNU/Linux a program called 'linitian' is used to make sure
that packages conforms to the Debian GNU/Linux packaging guide lines.
This fixes the problem reported as:
W: zfsutils: manpage-has-bad-whatis-entry usr/share/man/man1/zhack.1.gz
W: zfsutils: manpage-has-bad-whatis-entry usr/share/man/man8/fsck.zfs.8.gz
Not something that ZoL needs to addhere to, but every other man page
have their NAME section in a special way - why not these two as well?
Signed-off-by: Turbo Fredriksson <turbo@bayour.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2161
4088 use after free in arc_release()
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Garrett D'Amore <garrett@damore.org>
Reviewed by: Saso Kiselkov <skiselkov.ml@gmail.com>
Approved by: Dan McDonald <danmcd@nexenta.com>
References:
https://www.illumos.org/issues/4088illumos/illumos-gate@ccc22e1304
From the illumos issue:
A race-induced use after free occurs in arc_release() where the
ARC header is used outside the critical section protected by the
hash_lock.
Ported by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Richard Yao <ryao@gentoo.org>
Closes#2162
The spa_label_features nvlist is used in the sync context during pool
version upgrade.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2168
These are needed by consumers (i.e. Lustre) who wish to perform a
callback in the syncing context. Both a blocking and non-blocking
version are available to callers.
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Some callers of dmu_tx_assign() use the TXG_NOWAIT flag and call
dmu_tx_wait() themselves before retrying if the assignment fails.
The wait times for such callers are not accounted for in the
dmu_tx_assign kstat histogram, because the histogram only records
time spent in dmu_tx_assign(). This change moves the histogram
update to dmu_tx_wait() to properly account for all time spent there.
One downside of this approach is that it is possible to call
dmu_tx_wait() multiple times before successfully assigning a
transaction, in which case the cumulative wait time would not be
recorded. However, this case should not often arise in practice,
because most callers currently use one of these forms:
dmu_tx_assign(tx, TXG_WAIT);
dmu_tx_assign(tx, waited ? TXG_WAITED : TXG_NOWAIT);
The first form should make just one call to dmu_tx_delay() inside of
dmu_tx_assign(). The second form retries with TXG_WAITED if the first
assignment fails and incurs a delay, in which case no further waiting
is performed. Therefore transaction delays normally occur in one
call to dmu_tx_wait() so the histogram should be fairly accurate.
Another possible downside of this approach is that the histogram will
no longer record overhead outside of dmu_tx_wait() such as in
dmu_tx_try_assign(). While I'm not aware of any reason for concern on
this point, it is conceivable that lock contention, long list
traversal, etc. could cause assignment delays that would not be
reflected in the histogram. Therefore the histogram should strictly
be used for visibility in to the normal delay mechanisms and not as a
profiling tool for code performance.
Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1915
The nreserved column in the txgs kstat file always contains 0
following the write throttle restructuring of commit
e8b96c6007.
Prior to that commit, the nreserved column showed the number of bytes
temporarily reserved in the pool by a transaction group at sync time.
The new write throttle did away with temporary reservations and uses
the amount of dirty data instead. To approximate the old output of
the txgs kstat, the number of dirty bytes per-txg was passed in as
the nreserved value to spa_txg_history_set_io(). This approach did
not work as intended, because the per-txg dirty value is decremented
as data is written out to disk, so it is zero by the time we call
spa_txg_history_set_io(). To fix this, save the number of dirty
bytes before calling spa_sync(), and pass this value in to
spa_txg_history_set_io().
Also, since the name "nreserved" is now a misnomer, the column
heading is now labeled "ndirty".
Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #1696
A few counters in the dmu_tx kstats are obsolete or no longer
bumped properly.
- The sync task restructuring commit
13fe019870 removed the code
that bumpted dmu_tx_quota. The counter is now bumped in two
cases, instead of just the one case as before (after the result
of dsl_dataset_check_quota call). The second case is where
we check the requested reservation against the actual pool size,
as this is an implicit quota of sorts.
- The write throttle restructuring commit
e8b96c6007 makes dmu_tx_how and
dmu_tx_inflight obsolete, so they are removed.
Signed-off-by: Kohsuke Kawaguchi <kk@kohsuke.org>
Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#1914
Userland tools such as blkid, grub2-probe and zdb will go through the
buffer cache. However, ZFS uses on submit_bio() to bypass the buffer
cache when performing IO operations on vdevs for efficiency purposes.
This permits the on-disk state and buffer cache to fall out of
synchronization. That causes seemingly random failures when tools
reading stale metadata from the buffer cache try to access references to
data that is no longer there.
A particularly bad failure this causes involves grub2-probe, which is
used by grub2-mkconfig. Ordinarily, a rootfs might be called
rpool/ROOT/gentoo. However, when a failure occurs in grub2-probe,
grub2-mkconfig will generate a configuration file containing
/ROOT/gentoo, which omits the pool name and causes a boot failure.
This is avoidable by calling invalidate_bdev() on each flush, which is a
simple way to ensure that all non-dirty pages are wiped. Since userland
tools rarely access vdevs directly, this should be a fancy noop >99.999%
of the time and have little impact on IO. We could have tried a finer
grained approach for the rare instances in which the vdevs are accessed
frequently by userland. However, that would require consideration of
corner cases and it is not worth the effort.
Memory-wise, it would have been better to use a Linux kernel API hook to
disable the buffer cache on such devices, but it provides us no way of
doing that, so we opt for this approach instead. We should revisit that
idea in the future when higher priority issues have been tackled.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2150
p_l in #zfsonlinux reported that he had issues mounting filesystems that
were resolved by adding rc_need="mtab" to /etc/init.d/zfs. Closer
inspection revealed that we do have a race, but it is not clear how this
race caused mounting to fail. What is clear is that this race should be
fixed, so lets add the proper `use mtab` line to handle it.
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2148
4574 get_clones_stat does not call zap_count in non-debug kernel
Reviewed by: Matthew Ahrens <mahrens@delphix.com>
Reviewed by: Marcel Telka <marcel@telka.sk>
Approved by: Gordon Ross <gwr@nexenta.com>
References:
https://www.illumos.org/issues/4574illumos/illumos-gate@03d1795fa6
Ported-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2147
The length (number of integers) argument passed to zap_lookup was wrong;
likely as a result of performing stack-reduction on the function.
Signed-off-by: Tim Chase <tim@chase2k.com>
Signed-off-by: Richard Yao <ryao@gentoo.org>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2141
Remove recursion from dsl_dir_willuse_space() to reduce stack usage.
Issues with stack overflow were observed in zfs recv of zvols,
likelihood of an overflow is proportional to the depth of the dataset
as dsl_dir_willuse_space() recurses to parent datasets.
Signed-off-by: Andrew Barnes <barnes333@gmail.com>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2069
This stack of patches has been empirically shown to drastically improve
the hit rate of the ARC for certain workloads. As a result, fewer reads
to disk are required, which is generally a good thing and can
drastically improve performance if the workload is disk limited.
For the impatient, I'll summarize the results of the tests performed:
* Test 1 - Creating many empty directories. This test saw 99.9%
fewer reads and 12.8% more inodes created when running
*with* these changes.
* Test 2 - Creating many empty files. This test saw 4% fewer reads
and 0% more inodes created when running *with* these
changes.
* Test 3 - Creating many 4 KiB files. This test saw 96.7% fewer
reads and 4.9% more inodes created when running *with*
these changes.
* Test 4 - Creating many 4096 KiB files. This test saw 99.4% fewer
reads and 0% more inodes created (but took 6.9% fewer
seconds to complete) when running *with* these changes.
* Test 5 - Rsync'ing a dataset with many empty directories. This
test saw 36.2% fewer reads and 66.2% more inodes created
when running *with* these changes.
* Test 6 - Rsync'ing a dataset with many empty files. This test saw
30.9% fewer reads and 0% more inodes created (but took
24.3% fewer seconds to complete) when running *with*
these changes.
* Test 7 - Rsync'ing a dataset with many 4 KiB files. This test saw
30.8% fewer reads and 173.3% more inodes created when
running *with* these changes.
For the patient, the following consists of more a more detailed
description of the tests performed and the results gathered.
All the tests were run using identical machines, each with a pool
consisting of 5 mirror pairs with 2TB 7200 RPM disks. Each test was run
twice, once *without* this set of patches and again *with* this set of
patches to highlight the performance changes introduced. The first four
workloads tested were:
** NOTE: None of these tests were run to completion. They ran for a
set amount of time and then were terminated or hit ENOSPC.
1. Creating many empty directories:
* fdtree -d 10 -l 8 -s 0 -f 0 -C
-> 111,111,111 Directories
-> 0 Files
-> 0 KiB File Data
2. Creating many empty files:
* fdtree -d 10 -l 5 -s 0 -f 10000 -C
-> 111,111 Directories
-> 1,111,110,000 Files
-> 0 KiB File Data
3. Creating many 4 KiB files:
* fdtree -d 10 -l 5 -s 1 -f 10000 -C
-> 111,111 Directories
-> 1,111,110,000 Files
-> 4,444,440,000 KiB File Data
4. Creating many 4096 KiB files:
* fdtree -d 10 -l 5 -s 1024 -f 10000 -C
-> 111,111 Directories
-> 1,111,110,000 Files
-> 4,551,106,560,000 KiB File Data
Results for these first four tests are below:
| Time (s) | inodes | reads | writes |
--+----------+----------+--------+-----------+
Test 1 Before | 65069 | 37845363 | 831975 | 3214646 |
Test 1 After | 65069 | 42703608 | 778 | 3327674 |
--+----------+----------+--------+-----------+
Test 2 Before | 65073 | 54257583 | 208647 | 2413056 |
Test 2 After | 65069 | 54255782 | 200038 | 2533759 |
--+----------+----------+--------+-----------+
Test 3 Before | 65068 | 49857744 | 487130 | 5533348 |
Test 3 After | 65071 | 52294311 | 16078 | 5648354 |
--+----------+----------+--------+-----------+
Test 4 Before | 34854 | 2448329 | 385870 | 162116572 |
Test 4 After | 32419 | 2448329 | 2339 | 162175706 |
--+----------+----------+--------+-----------+
* "Time (s)" - The run time of the test in seconds
* "inodes" - The number of inodes created by the test
* "reads" - The number of reads performed by the test
* "writes" - The number of writes performed by the test
As you can see from the table above, running with this patch stack
*significantly* reduced the number of reads performed in 3 out of the 4
tests (due to an improved ARC hit rate).
In addition to the tests described above, which specifically targeted
creates only, three other workloads were tested. These additional tests
were targeting rsync performance against the datasets created in the
previous tests. A brief description of the workloads and results for
these tests are below:
** NOTE: Aside from (6), these tests didn't run to completion. They
ran for a set amount of time and then were terminated.
5. Rsync the dataset created in Test 1 to a new dataset:
* rsync -a /tank/test-1 /tank/test-5
6. Rsync the dataset created in Test 2 to a new dataset:
* rsync -a /tank/test-2 /tank/test-6
7. Rsync the dataset created in Test 3 to a new dataset:
* rsync -a /tank/test-3 /tank/test-7
Results for Test 5, 6, and 7 are below:
| Time (s) | inodes | reads | writes |
--+----------+----------+----------+---------+
Test 5 Before | 93041 | 17921014 | 47632823 | 4094848 |
Test 5 After | 93029 | 29785847 | 30376206 | 4484459 |
--+----------+----------+----------+---------+
Test 6 Before | 15290 | 54264474 | 6018331 | 733087 |
Test 6 After | 11573 | 54260826 | 4155661 | 617285 |
--+----------+----------+----------+---------+
Test 7 Before | 93057 | 10093749 | 41561635 | 3659098 |
Test 7 After | 93045 | 27587043 | 28773151 | 5612234 |
--+----------+----------+----------+---------+
* "Time (s)" - The run time of the test in seconds
* "inodes" - The number of inodes created by the test
* "reads" - The number of reads performed by the test
* "writes" - The number of writes performed by the test
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes#2110
Unfortunately, this change is an cheap attempt to work around a
pathological workload for the ARC. A "real" solution still needs to be
fleshed out, so this patch is intended to alleviate the situation in the
meantime. Let me try and describe the problem..
Data buffers residing in the dbuf hash table (dbuf cache) will keep a
hold on their respective dnode, this dnode will in turn keep a hold on
its backing dbuf (the physical block of the dnode object backing it).
Since the dnode has a hold on its backing dbuf, the arc buffer for this
dbuf is unevictable. What this essentially boils down to, "data" buffers
have the potential to pin "metadata" in the arc (as a result of these
dnode object buffers being unevictable).
This scenario becomes a real problem when the workload consists of many
small files (e.g. creating millions of 4K files). With this workload,
the arc's "arc_meta_used" space get filled up with buffers for any
resident directories as well as buffers for the objset's dnode object.
Once the "arc_meta_limit" is reached, the directory buffers will be
evicted and only the unevictable dnode object buffers will reside. If
the workload is simply creating new small files, these dnode object
buffers will never even be needed again, whereas the directory buffers
will be used constantly until the creates move to a new directory.
If "arc_c" and "arc_meta_limit" are sized appropriately, this
situation wont occur. This is because as the data buffers accumulate,
"arc_size" will eventually approach "arc_c" (before "arc_meta_used"
reaches "arc_meta_limit"); at that point the data buffers will be
evicted, which releases the hold on the dnode, which releases the hold
on the dnode object's dbuf, which allows that buffer to be evicted from
the arc in preference to more "useful" metadata.
So, to side step the issue, we simply need to ensure "arc_size" reaches
"arc_c" before "arc_meta_used" reaches "arc_meta_limit". In order to
pick a proper limit, we have to do some math.
To make things a little easier to follow, it is assumed that there will
only be a single data buffer per file (which is probably always the case
for "small" files anyways).
Based on the current internals of the arc, if N files residing in the
dbuf cache all pin a single dnode buffer (i.e. their dnodes all share
the same physical dnode object block), then the following amount of
"arc_meta_used" space will be consumed:
- 16K for the dnode object's block - [ 16384 bytes]
- N * sizeof(dnode_t) -------------- [ N * 928 bytes]
- (N + 1) * sizeof(arc_buf_t) ------ [(N + 1) * 72 bytes]
- (N + 1) * sizeof(arc_buf_hdr_t) -- [(N + 1) * 264 bytes]
- (N + 1) * sizeof(dmu_buf_impl_t) - [(N + 1) * 280 bytes]
To simplify, these N files will pin the following amount of
"arc_meta_used" space as unevictable:
Pinned "arc_meta_used" bytes = 16384 + N * 928 + (N + 1) * (72 + 264 + 280)
Pinned "arc_meta_used" bytes = 17000 + N * 1544
This pinned space is regardless of the size of the files, and is only
dependent on the number of pinned dnodes sharing a physical block
(i.e. N). For example, 32 512b files sharing a single dnode object
block would consume the same "arc_meta_used" space as 32 4K files
sharing a single dnode object block.
Now, given a files size of S, we can determine the total amount of
space that will be consumed in the arc:
Total = 17000 + N * 1544 + S * N
^^^^^^^^^^^^^^^^ ^^^^^
metadata data
So, given these formulas, we can generate a table which states the ratio
of pinned metadata to total arc (meta + data) using different values of
N (number of pinned dnodes per pinned physical dnode block) and S (size
of the file).
File Sizes (S)
| 512 | 1024 | 2048 | 4096 | 8192 | 16384 |
---+----------+----------+----------+----------+----------+----------+
1 | 0.973132 | 0.947670 | 0.900544 | 0.819081 | 0.693597 | 0.530921 |
2 | 0.951497 | 0.907481 | 0.830632 | 0.710325 | 0.550779 | 0.380051 |
N 4 | 0.918807 | 0.849809 | 0.738842 | 0.585844 | 0.414271 | 0.261250 |
8 | 0.877541 | 0.781803 | 0.641770 | 0.472505 | 0.309333 | 0.182965 |
16 | 0.835819 | 0.717945 | 0.559996 | 0.388885 | 0.241376 | 0.137253 |
32 | 0.802106 | 0.669597 | 0.503304 | 0.336277 | 0.202123 | 0.112423 |
As you can see, if we wanted to support the absolute worst case of 1
dnode per physical dnode block and 512b files, we would have to set the
"arc_meta_limit" to something greater than 97.3132% of "arc_c_max". At
that point, it essentially defeats the purpose of having an
"arc_meta_limit" at all.
This patch changes the default value of "arc_meta_limit" to be 75% of
"arc_c_max", which should be good enough for "most" workloads (I think).
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
Previously, the "data_size" field in the arcstats kstat contained the
amount of cached "metadata" and "data" in the ARC. The problem is this
then made it difficult to extract out just the "metadata" size, or just
the "data" size.
To make it easier to distinguish the two values, "data_size" has been
modified to count only buffers of type ARC_BUFC_DATA, and "meta_size"
was added to count only buffers of type ARC_BUFC_METADATA. If one wants
the old "data_size" value, simply sum the new "data_size" and
"meta_size" values.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
When the arc is at it's size limit and a new buffer is added, data will
be evicted (or recycled) from the arc to make room for this new buffer.
As far as I can tell, this is to try and keep the arc from over stepping
it's bounds (i.e. keep it below the size limitation placed on it).
This makes sense conceptually, but there appears to be a subtle flaw in
its current implementation, resulting in metadata buffers being
throttled. When it evicts from the arc's lists, it also passes in a
"type" so as to remove a buffer of the same type that it is adding. The
problem with this is that once the size limit is hit, the ratio of
"metadata" to "data" contained in the arc essentially becomes fixed.
For example, consider the following scenario:
* the size of the arc is capped at 10G
* the meta_limit is capped at 4G
* 9G of the arc contains "data"
* 1G of the arc contains "metadata"
Now, every time a new "metadata" buffer is created and added to the arc,
an older "metadata" buffer(s) will be removed from the arc; preserving
the 9G "data" to 1G "metadata" ratio that was in-place when the size
limit was reached. This occurs even though the amount of "metadata" is
far below the "metadata" limit. This can result in the arc behaving
pathologically for certain workloads.
To fix this, the arc_get_data_buf function was modified to evict "data"
from the arc even when adding a "metadata" buffer; unless it's at the
"metadata" limit. In addition, arc_evict now more closely resembles
arc_evict_ghost; such that when evicting "data" from the arc, it may
make a second pass over the arc lists and evict "metadata" if it cannot
meet the eviction size the first time around.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
Using "arc_meta_used" to determine if the arc's mru list is over it's
target value of "arc_p" doesn't seem correct. The size of the mru list
and the value of "arc_meta_used", although related, are completely
independent. Buffers contained in "arc_meta_used" may not even be
contained in the arc's mru list. As such, this patch removes
"arc_meta_used" from the calculation in arc_adjust.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
To maintain a strict limit on the metadata contained in the arc, while
preventing the arc buffer headers from completely consuming the
"arc_meta_used" space, we need to evict metadata buffers from the arc's
ghost lists along with the regular lists.
This change modifies arc_adjust_meta such that it more closely models
the adjustments made in arc_adjust. "arc_meta_used" is used similarly to
"arc_size", and "arc_meta_limit" is used similarly to "arc_c".
Testing metadata intensive workloads (e.g. creating, copying, and
removing millions of small files and/or directories) has shown this
change to make a dramatic improvement to the hit rate maintained in the
arc. While I think there is still room for improvement, this is a big
step in the right direction.
In addition, zpl_free_cached_objects was made into a no-op as I'm not
yet sure how to properly implement that function.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
This reverts commit c11a12bc3b.
Out of memory events were fixed by reverting this patch.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
It's unclear why adjustments to arc_p need to be dampened as they are in
arc_adjust. With that said, it's removal significantly improves the arc's
ability to "warm up" to a given workload. Thus, I'm disabling by default
until its usefulness is better understood.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
Setting a limit on the minimum value of "arc_p" has been shown to have
detrimental effects on the arc hit rate for certain "metadata" intensive
workloads. Specifically, this has been exhibited with a workload that
constantly dirties new "metadata" but also frequently touches a "small"
amount of mfu data (e.g. mkdir's).
What is seen is that the new anon data throttles the mfu list to a
negligible size (because arc_p > anon + mru in arc_get_data_buf), even
though the mfu ghost list receives a constant stream of hits. To remedy
this, arc_p is now allowed to drop to zero if the algorithm deems it
necessary.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
For specific workloads consisting mainly of mfu data and new anon data
buffers, the aggressive growth of arc_p found in the arc_get_data_buf()
function can have detrimental effects on the mfu list size and ghost
list hit rate.
Running a workload consisting of two processes:
* Process 1 is creating many small files
* Process 2 is tar'ing a directory consisting of many small files
I've seen arc_p and the mru grow to their maximum size, while the mru
ghost list receives 100K times fewer hits than the mfu ghost list.
Ideally, as the mfu ghost list receives hits, arc_p should be driven
down and the size of the mfu should increase. Given the specific
workload I was testing with, the mfu list size should grow to a point
where almost no mfu ghost list hits would occur. Unfortunately, this
does not happen because the newly dirtied anon buffers constancy drive
arc_p to its maximum value and keep it there (effectively prioritizing
the mru list and starving the mfu list down to a negligible size).
The logic to increment arc_p from within the arc_get_data_buf() function
was introduced many years ago in this upstream commit:
commit 641fbdae3a027d12b3c3dcd18927ccafae6d58bc
Author: maybee <none@none>
Date: Wed Dec 20 15:46:12 2006 -0800
6505658 target MRU size (arc.p) needs to be adjusted more aggressively
and since I don't fully understand the motivation for the change, I am
reluctant to completely remove it.
As a way to test out how it's removal might affect performance, I've
disabled that code by default, but left it tunable via a module option.
Thus, if its removal is found to be grossly detrimental for certain
workloads, it can be re-enabled on the fly, without a code change.
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
In an attempt to prevent arc_c from collapsing "too fast", the
arc_shrink() function was updated to take a "bytes" parameter by this
change:
commit 302f753f16
Author: Brian Behlendorf <behlendorf1@llnl.gov>
Date: Tue Mar 13 14:29:16 2012 -0700
Integrate ARC more tightly with Linux
Unfortunately, that change failed to make a similar change to the way
that arc_p was updated. So, there still exists the possibility for arc_p
to collapse to near 0 when the kernel start calling the arc's shrinkers.
This change attempts to fix this, by decrementing arc_p by the "bytes"
parameter in the same way that arc_c is updated.
In addition, a minimum value of arc_p is attempted to be maintained,
similar to the way a minimum arc_p value is maintained in arc_adapt().
Signed-off-by: Prakash Surya <surya1@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Issue #2110
Brian Behlendorf
Chris Dunlap
Gunnar Beutner
Richard Yao
Boris Protopopov
Yuri Pankov
Chunwei Chen
Isaac Huang
Andrey Vesnovaty
Turbo Fredriksson
Tim Chase
Ned Bass
Alexander Stetsenko
cburroughs
Andrew Barnes
Prakash Surya