c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2024-11-17 10:01:01 +03:00
Code Issues Packages Projects Releases Wiki Activity
f6fbe25664
mirror_zfs/man/man8/zdb.8
Serapheim Dimitropoulos 93e28d661e Log Spacemap Project 
= Motivation

At Delphix we've seen a lot of customer systems where fragmentation
is over 75% and random writes take a performance hit because a lot
of time is spend on I/Os that update on-disk space accounting metadata.
Specifically, we seen cases where 20% to 40% of sync time is spend
after sync pass 1 and ~30% of the I/Os on the system is spent updating
spacemaps.

The problem is that these pools have existed long enough that we've
touched almost every metaslab at least once, and random writes
scatter frees across all metaslabs every TXG, thus appending to
their spacemaps and resulting in many I/Os. To give an example,
assuming that every VDEV has 200 metaslabs and our writes fit within
a single spacemap block (generally 4K) we have 200 I/Os. Then if we
assume 2 levels of indirection, we need 400 additional I/Os and
since we are talking about metadata for which we keep 2 extra copies
for redundancy we need to triple that number, leading to a total of
1800 I/Os per VDEV every TXG.

We could try and decrease the number of metaslabs so we have less
I/Os per TXG but then each metaslab would cover a wider range on
disk and thus would take more time to be loaded in memory from disk.
In addition, after it's loaded, it's range tree would consume more
memory.

Another idea would be to just increase the spacemap block size
which would allow us to fit more entries within an I/O block
resulting in fewer I/Os per metaslab and a speedup in loading time.
The problem is still that we don't deal with the number of I/Os
going up as the number of metaslabs is increasing and the fact
is that we generally write a lot to a few metaslabs and a little
to the rest of them. Thus, just increasing the block size would
actually waste bandwidth because we won't be utilizing our bigger
block size.

= About this patch

This patch introduces the Log Spacemap project which provides the
solution to the above problem while taking into account all the
aforementioned tradeoffs. The details on how it achieves that can
be found in the references sections below and in the code (see
Big Theory Statement in spa_log_spacemap.c).

Even though the change is fairly constraint within the metaslab
and lower-level SPA codepaths, there is a side-change that is
user-facing. The change is that VDEV IDs from VDEV holes will no
longer be reused. To give some background and reasoning for this,
when a log device is removed and its VDEV structure was replaced
with a hole (or was compacted; if at the end of the vdev array),
its vdev_id could be reused by devices added after that. Now
with the pool-wide space maps recording the vdev ID, this behavior
can cause problems (e.g. is this entry referring to a segment in
the new vdev or the removed log?). Thus, to simplify things the
ID reuse behavior is gone and now vdev IDs for top-level vdevs
are truly unique within a pool.

= Testing

The illumos implementation of this feature has been used internally
for a year and has been in production for ~6 months. For this patch
specifically there don't seem to be any regressions introduced to
ZTS and I have been running zloop for a week without any related
problems.

= Performance Analysis (Linux Specific)

All performance results and analysis for illumos can be found in
the links of the references. Redoing the same experiments in Linux
gave similar results. Below are the specifics of the Linux run.

After the pool reached stable state the percentage of the time
spent in pass 1 per TXG was 64% on average for the stock bits
while the log spacemap bits stayed at 95% during the experiment
(graph: sdimitro.github.io/img/linux-lsm/PercOfSyncInPassOne.png).

Sync times per TXG were 37.6 seconds on average for the stock
bits and 22.7 seconds for the log spacemap bits (related graph:
sdimitro.github.io/img/linux-lsm/SyncTimePerTXG.png). As a result
the log spacemap bits were able to push more TXGs, which is also
the reason why all graphs quantified per TXG have more entries for
the log spacemap bits.

Another interesting aspect in terms of txg syncs is that the stock
bits had 22% of their TXGs reach sync pass 7, 55% reach sync pass 8,
and 20% reach 9. The log space map bits reached sync pass 4 in 79%
of their TXGs, sync pass 7 in 19%, and sync pass 8 at 1%. This
emphasizes the fact that not only we spend less time on metadata
but we also iterate less times to convergence in spa_sync() dirtying
objects.
[related graphs:
stock- sdimitro.github.io/img/linux-lsm/NumberOfPassesPerTXGStock.png
lsm- sdimitro.github.io/img/linux-lsm/NumberOfPassesPerTXGLSM.png]

Finally, the improvement in IOPs that the userland gains from the
change is approximately 40%. There is a consistent win in IOPS as
you can see from the graphs below but the absolute amount of
improvement that the log spacemap gives varies within each minute
interval.
sdimitro.github.io/img/linux-lsm/StockVsLog3Days.png
sdimitro.github.io/img/linux-lsm/StockVsLog10Hours.png

= Porting to Other Platforms

For people that want to port this commit to other platforms below
is a list of ZoL commits that this patch depends on:

Make zdb results for checkpoint tests consistent
db587941c5

Update vdev_is_spacemap_addressable() for new spacemap encoding
419ba59145

Simplify spa_sync by breaking it up to smaller functions
8dc2197b7b

Factor metaslab_load_wait() in metaslab_load()
b194fab0fb

Rename range_tree_verify to range_tree_verify_not_present
df72b8bebe

Change target size of metaslabs from 256GB to 16GB
c853f382db

zdb -L should skip leak detection altogether
21e7cf5da8

vs_alloc can underflow in L2ARC vdevs
7558997d2f

Simplify log vdev removal code
6c926f426a

Get rid of space_map_update() for ms_synced_length
425d3237ee

Introduce auxiliary metaslab histograms
928e8ad47d

Error path in metaslab_load_impl() forgets to drop ms_sync_lock
8eef997679

= References

Background, Motivation, and Internals of the Feature
- OpenZFS 2017 Presentation:
youtu.be/jj2IxRkl5bQ
- Slides:
slideshare.net/SerapheimNikolaosDim/zfs-log-spacemaps-project

Flushing Algorithm Internals & Performance Results
(Illumos Specific)
- Blogpost:
sdimitro.github.io/post/zfs-lsm-flushing/
- OpenZFS 2018 Presentation:
youtu.be/x6D2dHRjkxw
- Slides:
slideshare.net/SerapheimNikolaosDim/zfs-log-spacemap-flushing-algorithm

Upstream Delphix Issues:
DLPX-51539, DLPX-59659, DLPX-57783, DLPX-61438, DLPX-41227, DLPX-59320
DLPX-63385

Reviewed-by: Sean Eric Fagan <sef@ixsystems.com>
Reviewed-by: Matt Ahrens <matt@delphix.com>
Reviewed-by: George Wilson <gwilson@delphix.com>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Signed-off-by: Serapheim Dimitropoulos <serapheim@delphix.com>
Closes #8442
2019-07-16 10:11:49 -07:00

422 lines
12 KiB
Groff
Raw Blame History

.\"
.\" This file and its contents are supplied under the terms of the
.\" Common Development and Distribution License ("CDDL"), version 1.0.
.\" You may only use this file in accordance with the terms of version
.\" 1.0 of the CDDL.
.\"
.\" A full copy of the text of the CDDL should have accompanied this
.\" source. A copy of the CDDL is also available via the Internet at
.\" http://www.illumos.org/license/CDDL.
.\"
.\"
.\" Copyright 2012, Richard Lowe.
.\" Copyright (c) 2012, 2018 by Delphix. All rights reserved.
.\" Copyright 2017 Nexenta Systems, Inc.
.\" Copyright (c) 2017 Lawrence Livermore National Security, LLC.
.\" Copyright (c) 2017 Intel Corporation.
.\"
.Dd April 14, 2019
.Dt ZDB 8 SMM
.Os Linux
.Sh NAME
.Nm zdb
.Nd display zpool debugging and consistency information
.Sh SYNOPSIS
.Nm
.Op Fl AbcdDFGhikLMPsvXY
.Op Fl e Oo Fl V Oc Op Fl p Ar path ...
.Op Fl I Ar inflight I/Os
.Oo Fl o Ar var Ns = Ns Ar value Oc Ns ...
.Op Fl t Ar txg
.Op Fl U Ar cache
.Op Fl x Ar dumpdir
.Op Ar poolname Op Ar object ...
.Nm
.Op Fl AdiPv
.Op Fl e Oo Fl V Oc Op Fl p Ar path ...
.Op Fl U Ar cache
.Ar dataset Op Ar object ...
.Nm
.Fl C
.Op Fl A
.Op Fl U Ar cache
.Nm
.Fl E
.Op Fl A
.Ar word0 Ns \&: Ns Ar word1 Ns :...: Ns Ar word15
.Nm
.Fl l
.Op Fl Aqu
.Ar device
.Nm
.Fl m
.Op Fl AFLPXY
.Op Fl e Oo Fl V Oc Op Fl p Ar path ...
.Op Fl t Ar txg
.Op Fl U Ar cache
.Ar poolname Op Ar vdev Op Ar metaslab ...
.Nm
.Fl O
.Ar dataset path
.Nm
.Fl R
.Op Fl A
.Op Fl e Oo Fl V Oc Op Fl p Ar path ...
.Op Fl U Ar cache
.Ar poolname vdev Ns \&: Ns Ar offset Ns \&: Ns Ar size Ns Op : Ns Ar flags
.Nm
.Fl S
.Op Fl AP
.Op Fl e Oo Fl V Oc Op Fl p Ar path ...
.Op Fl U Ar cache
.Ar poolname
.Sh DESCRIPTION
The
.Nm
utility displays information about a ZFS pool useful for debugging and performs
some amount of consistency checking.
It is a not a general purpose tool and options
.Pq and facilities
may change.
This is not a
.Xr fsck 8
utility.
.Pp
The output of this command in general reflects the on-disk structure of a ZFS
pool, and is inherently unstable.
The precise output of most invocations is not documented, a knowledge of ZFS
internals is assumed.
.Pp
If the
.Ar dataset
argument does not contain any
.Qq Sy /
or
.Qq Sy @
characters, it is interpreted as a pool name.
The root dataset can be specified as
.Ar pool Ns /
.Pq pool name followed by a slash .
.Pp
When operating on an imported and active pool it is possible, though unlikely,
that zdb may interpret inconsistent pool data and behave erratically.
.Sh OPTIONS
Display options:
.Bl -tag -width Ds
.It Fl b
Display statistics regarding the number, size
.Pq logical, physical and allocated
and deduplication of blocks.
.It Fl c
Verify the checksum of all metadata blocks while printing block statistics
.Po see
.Fl b
.Pc .
.Pp
If specified multiple times, verify the checksums of all blocks.
.It Fl C
Display information about the configuration.
If specified with no other options, instead display information about the cache
file
.Pq Pa /etc/zfs/zpool.cache .
To specify the cache file to display, see
.Fl U .
.Pp
If specified multiple times, and a pool name is also specified display both the
cached configuration and the on-disk configuration.
If specified multiple times with
.Fl e
also display the configuration that would be used were the pool to be imported.
.It Fl d
Display information about datasets.
Specified once, displays basic dataset information: ID, create transaction,
size, and object count.
.Pp
If specified multiple times provides greater and greater verbosity.
.Pp
If object IDs are specified, display information about those specific objects
only.
.It Fl D
Display deduplication statistics, including the deduplication ratio
.Pq Sy dedup ,
compression ratio
.Pq Sy compress ,
inflation due to the zfs copies property
.Pq Sy copies ,
and an overall effective ratio
.Pq Sy dedup No * Sy compress No / Sy copies .
.It Fl DD
Display a histogram of deduplication statistics, showing the allocated
.Pq physically present on disk
and referenced
.Pq logically referenced in the pool
block counts and sizes by reference count.
.It Fl DDD
Display the statistics independently for each deduplication table.
.It Fl DDDD
Dump the contents of the deduplication tables describing duplicate blocks.
.It Fl DDDDD
Also dump the contents of the deduplication tables describing unique blocks.
.It Fl E Ar word0 Ns \&: Ns Ar word1 Ns :...: Ns Ar word15
Decode and display block from an embedded block pointer specified by the
.Ar word
arguments.
.It Fl h
Display pool history similar to
.Nm zpool Cm history ,
but include internal changes, transaction, and dataset information.
.It Fl i
Display information about intent log
.Pq ZIL
entries relating to each dataset.
If specified multiple times, display counts of each intent log transaction type.
.It Fl k
Examine the checkpointed state of the pool.
Note, the on disk format of the pool is not reverted to the checkpointed state.
.It Fl l Ar device
Read the vdev labels from the specified device.
.Nm Fl l
will return 0 if valid label was found, 1 if error occurred, and 2 if no valid
labels were found. Each unique configuration is displayed only once.
.It Fl ll Ar device
In addition display label space usage stats.
.It Fl lll Ar device
Display every configuration, unique or not.
.Pp
If the
.Fl q
option is also specified, don't print the labels.
.Pp
If the
.Fl u
option is also specified, also display the uberblocks on this device. Specify
multiple times to increase verbosity.
.It Fl L
Disable leak detection and the loading of space maps.
By default,
.Nm
verifies that all non-free blocks are referenced, which can be very expensive.
.It Fl m
Display the offset, spacemap, free space of each metaslab, all the log
spacemaps and their obsolete entry statistics.
.It Fl mm
Also display information about the on-disk free space histogram associated with
each metaslab.
.It Fl mmm
Display the maximum contiguous free space, the in-core free space histogram, and
the percentage of free space in each space map.
.It Fl mmmm
Display every spacemap record.
.It Fl M
Display the offset, spacemap, and free space of each metaslab.
.It Fl MM
Also display information about the maximum contiguous free space and the
percentage of free space in each space map.
.It Fl MMM
Display every spacemap record.
.It Fl O Ar dataset path
Look up the specified
.Ar path
inside of the
.Ar dataset
and display its metadata and indirect blocks.
Specified
.Ar path
must be relative to the root of
.Ar dataset .
This option can be combined with
.Fl v
for increasing verbosity.
.It Xo
.Fl R Ar poolname vdev Ns \&: Ns Ar offset Ns \&: Ns Ar size Ns Op : Ns Ar flags
.Xc
Read and display a block from the specified device.
By default the block is displayed as a hex dump, but see the description of the
.Sy r
flag, below.
.Pp
The block is specified in terms of a colon-separated tuple
.Ar vdev
.Pq an integer vdev identifier
.Ar offset
.Pq the offset within the vdev
.Ar size
.Pq the size of the block to read
and, optionally,
.Ar flags
.Pq a set of flags, described below .
.Pp
.Bl -tag -compact -width "b offset"
.It Sy b Ar offset
Print block pointer
.It Sy d
Decompress the block. Set environment variable
.Nm ZBD_NO_ZLE
to skip zle when guessing.
.It Sy e
Byte swap the block
.It Sy g
Dump gang block header
.It Sy i
Dump indirect block
.It Sy r
Dump raw uninterpreted block data
.El
.It Fl s
Report statistics on
.Nm zdb
I/O.
Display operation counts, bandwidth, and error counts of I/O to the pool from
.Nm .
.It Fl S
Simulate the effects of deduplication, constructing a DDT and then display
that DDT as with
.Fl DD .
.It Fl u
Display the current uberblock.
.El
.Pp
Other options:
.Bl -tag -width Ds
.It Fl A
Do not abort should any assertion fail.
.It Fl AA
Enable panic recovery, certain errors which would otherwise be fatal are
demoted to warnings.
.It Fl AAA
Do not abort if asserts fail and also enable panic recovery.
.It Fl e Op Fl p Ar path ...
Operate on an exported pool, not present in
.Pa /etc/zfs/zpool.cache .
The
.Fl p
flag specifies the path under which devices are to be searched.
.It Fl x Ar dumpdir
All blocks accessed will be copied to files in the specified directory.
The blocks will be placed in sparse files whose name is the same as
that of the file or device read.
.Nm
can be then run on the generated files.
Note that the
.Fl bbc
flags are sufficient to access
.Pq and thus copy
all metadata on the pool.
.It Fl F
Attempt to make an unreadable pool readable by trying progressively older
transactions.
.It Fl G
Dump the contents of the zfs_dbgmsg buffer before exiting
.Nm .
zfs_dbgmsg is a buffer used by ZFS to dump advanced debug information.
.It Fl I Ar inflight I/Os
Limit the number of outstanding checksum I/Os to the specified value.
The default value is 200.
This option affects the performance of the
.Fl c
option.
.It Fl o Ar var Ns = Ns Ar value ...
Set the given global libzpool variable to the provided value.
The value must be an unsigned 32-bit integer.
Currently only little-endian systems are supported to avoid accidentally setting
the high 32 bits of 64-bit variables.
.It Fl P
Print numbers in an unscaled form more amenable to parsing, eg. 1000000 rather
than 1M.
.It Fl t Ar transaction
Specify the highest transaction to use when searching for uberblocks.
See also the
.Fl u
and
.Fl l
options for a means to see the available uberblocks and their associated
transaction numbers.
.It Fl U Ar cachefile
Use a cache file other than
.Pa /etc/zfs/zpool.cache .
.It Fl v
Enable verbosity.
Specify multiple times for increased verbosity.
.It Fl V
Attempt verbatim import.
This mimics the behavior of the kernel when loading a pool from a cachefile.
Only usable with
.Fl e .
.It Fl X
Attempt
.Qq extreme
transaction rewind, that is attempt the same recovery as
.Fl F
but read transactions otherwise deemed too old.
.It Fl Y
Attempt all possible combinations when reconstructing indirect split blocks.
This flag disables the individual I/O deadman timer in order to allow as
much time as required for the attempted reconstruction.
.El
.Pp
Specifying a display option more than once enables verbosity for only that
option, with more occurrences enabling more verbosity.
.Pp
If no options are specified, all information about the named pool will be
displayed at default verbosity.
.Sh EXAMPLES
.Bl -tag -width Ds
.It Xo
.Sy Example 1
Display the configuration of imported pool
.Pa rpool
.Xc
.Bd -literal
# zdb -C rpool
MOS Configuration:
version: 28
name: 'rpool'
...
.Ed
.It Xo
.Sy Example 2
Display basic dataset information about
.Pa rpool
.Xc
.Bd -literal
# zdb -d rpool
Dataset mos [META], ID 0, cr_txg 4, 26.9M, 1051 objects
Dataset rpool/swap [ZVOL], ID 59, cr_txg 356, 486M, 2 objects
...
.Ed
.It Xo
.Sy Example 3
Display basic information about object 0 in
.Pa rpool/export/home
.Xc
.Bd -literal
# zdb -d rpool/export/home 0
Dataset rpool/export/home [ZPL], ID 137, cr_txg 1546, 32K, 8 objects
Object lvl iblk dblk dsize lsize %full type
0 7 16K 16K 15.0K 16K 25.00 DMU dnode
.Ed
.It Xo
.Sy Example 4
Display the predicted effect of enabling deduplication on
.Pa rpool
.Xc
.Bd -literal
# zdb -S rpool
Simulated DDT histogram:
bucket allocated referenced
______ ______________________________ ______________________________
refcnt blocks LSIZE PSIZE DSIZE blocks LSIZE PSIZE DSIZE
------ ------ ----- ----- ----- ------ ----- ----- -----
1 694K 27.1G 15.0G 15.0G 694K 27.1G 15.0G 15.0G
2 35.0K 1.33G 699M 699M 74.7K 2.79G 1.45G 1.45G
...
dedup = 1.11, compress = 1.80, copies = 1.00, dedup * compress / copies = 2.00
.Ed
.El
.Sh SEE ALSO
.Xr zfs 8 ,
.Xr zpool 8
Reference in New Issue View Git Blame Copy Permalink