c/mirror_zfs
1
0
Fork 0
mirror of https://git.proxmox.com/git/mirror_zfs.git synced 2024-12-27 19:39:35 +03:00
Code Issues Packages Projects Releases Wiki Activity
43e52eddb1
mirror_zfs/man/man5/zpool-features.5
Ned Bass 50c957f702 Implement large_dnode pool feature 
Justification
-------------

This feature adds support for variable length dnodes. Our motivation is
to eliminate the overhead associated with using spill blocks.  Spill
blocks are used to store system attribute data (i.e. file metadata) that
does not fit in the dnode's bonus buffer. By allowing a larger bonus
buffer area the use of a spill block can be avoided.  Spill blocks
potentially incur an additional read I/O for every dnode in a dnode
block. As a worst case example, reading 32 dnodes from a 16k dnode block
and all of the spill blocks could issue 33 separate reads. Now suppose
those dnodes have size 1024 and therefore don't need spill blocks.  Then
the worst case number of blocks read is reduced to from 33 to two--one
per dnode block. In practice spill blocks may tend to be co-located on
disk with the dnode blocks so the reduction in I/O would not be this
drastic. In a badly fragmented pool, however, the improvement could be
significant.

ZFS-on-Linux systems that make heavy use of extended attributes would
benefit from this feature. In particular, ZFS-on-Linux supports the
xattr=sa dataset property which allows file extended attribute data
to be stored in the dnode bonus buffer as an alternative to the
traditional directory-based format. Workloads such as SELinux and the
Lustre distributed filesystem often store enough xattr data to force
spill bocks when xattr=sa is in effect. Large dnodes may therefore
provide a performance benefit to such systems.

Other use cases that may benefit from this feature include files with
large ACLs and symbolic links with long target names. Furthermore,
this feature may be desirable on other platforms in case future
applications or features are developed that could make use of a
larger bonus buffer area.

Implementation
--------------

The size of a dnode may be a multiple of 512 bytes up to the size of
a dnode block (currently 16384 bytes). A dn_extra_slots field was
added to the current on-disk dnode_phys_t structure to describe the
size of the physical dnode on disk. The 8 bits for this field were
taken from the zero filled dn_pad2 field. The field represents how
many "extra" dnode_phys_t slots a dnode consumes in its dnode block.
This convention results in a value of 0 for 512 byte dnodes which
preserves on-disk format compatibility with older software.

Similarly, the in-memory dnode_t structure has a new dn_num_slots field
to represent the total number of dnode_phys_t slots consumed on disk.
Thus dn->dn_num_slots is 1 greater than the corresponding
dnp->dn_extra_slots. This difference in convention was adopted
because, unlike on-disk structures, backward compatibility is not a
concern for in-memory objects, so we used a more natural way to
represent size for a dnode_t.

The default size for newly created dnodes is determined by the value of
a new "dnodesize" dataset property. By default the property is set to
"legacy" which is compatible with older software. Setting the property
to "auto" will allow the filesystem to choose the most suitable dnode
size. Currently this just sets the default dnode size to 1k, but future
code improvements could dynamically choose a size based on observed
workload patterns. Dnodes of varying sizes can coexist within the same
dataset and even within the same dnode block. For example, to enable
automatically-sized dnodes, run

 # zfs set dnodesize=auto tank/fish

The user can also specify literal values for the dnodesize property.
These are currently limited to powers of two from 1k to 16k. The
power-of-2 limitation is only for simplicity of the user interface.
Internally the implementation can handle any multiple of 512 up to 16k,
and consumers of the DMU API can specify any legal dnode value.

The size of a new dnode is determined at object allocation time and
stored as a new field in the znode in-memory structure. New DMU
interfaces are added to allow the consumer to specify the dnode size
that a newly allocated object should use. Existing interfaces are
unchanged to avoid having to update every call site and to preserve
compatibility with external consumers such as Lustre. The new
interfaces names are given below. The versions of these functions that
don't take a dnodesize parameter now just call the _dnsize() versions
with a dnodesize of 0, which means use the legacy dnode size.

New DMU interfaces:
  dmu_object_alloc_dnsize()
  dmu_object_claim_dnsize()
  dmu_object_reclaim_dnsize()

New ZAP interfaces:
  zap_create_dnsize()
  zap_create_norm_dnsize()
  zap_create_flags_dnsize()
  zap_create_claim_norm_dnsize()
  zap_create_link_dnsize()

The constant DN_MAX_BONUSLEN is renamed to DN_OLD_MAX_BONUSLEN. The
spa_maxdnodesize() function should be used to determine the maximum
bonus length for a pool.

These are a few noteworthy changes to key functions:

* The prototype for dnode_hold_impl() now takes a "slots" parameter.
  When the DNODE_MUST_BE_FREE flag is set, this parameter is used to
  ensure the hole at the specified object offset is large enough to
  hold the dnode being created. The slots parameter is also used
  to ensure a dnode does not span multiple dnode blocks. In both of
  these cases, if a failure occurs, ENOSPC is returned. Keep in mind,
  these failure cases are only possible when using DNODE_MUST_BE_FREE.

  If the DNODE_MUST_BE_ALLOCATED flag is set, "slots" must be 0.
  dnode_hold_impl() will check if the requested dnode is already
  consumed as an extra dnode slot by an large dnode, in which case
  it returns ENOENT.

* The function dmu_object_alloc() advances to the next dnode block
  if dnode_hold_impl() returns an error for a requested object.
  This is because the beginning of the next dnode block is the only
  location it can safely assume to either be a hole or a valid
  starting point for a dnode.

* dnode_next_offset_level() and other functions that iterate
  through dnode blocks may no longer use a simple array indexing
  scheme. These now use the current dnode's dn_num_slots field to
  advance to the next dnode in the block. This is to ensure we
  properly skip the current dnode's bonus area and don't interpret it
  as a valid dnode.

zdb
---
The zdb command was updated to display a dnode's size under the
"dnsize" column when the object is dumped.

For ZIL create log records, zdb will now display the slot count for
the object.

ztest
-----
Ztest chooses a random dnodesize for every newly created object. The
random distribution is more heavily weighted toward small dnodes to
better simulate real-world datasets.

Unused bonus buffer space is filled with non-zero values computed from
the object number, dataset id, offset, and generation number.  This
helps ensure that the dnode traversal code properly skips the interior
regions of large dnodes, and that these interior regions are not
overwritten by data belonging to other dnodes. A new test visits each
object in a dataset. It verifies that the actual dnode size matches what
was stored in the ztest block tag when it was created. It also verifies
that the unused bonus buffer space is filled with the expected data
patterns.

ZFS Test Suite
--------------
Added six new large dnode-specific tests, and integrated the dnodesize
property into existing tests for zfs allow and send/recv.

Send/Receive
------------
ZFS send streams for datasets containing large dnodes cannot be received
on pools that don't support the large_dnode feature. A send stream with
large dnodes sets a DMU_BACKUP_FEATURE_LARGE_DNODE flag which will be
unrecognized by an incompatible receiving pool so that the zfs receive
will fail gracefully.

While not implemented here, it may be possible to generate a
backward-compatible send stream from a dataset containing large
dnodes. The implementation may be tricky, however, because the send
object record for a large dnode would need to be resized to a 512
byte dnode, possibly kicking in a spill block in the process. This
means we would need to construct a new SA layout and possibly
register it in the SA layout object. The SA layout is normally just
sent as an ordinary object record. But if we are constructing new
layouts while generating the send stream we'd have to build the SA
layout object dynamically and send it at the end of the stream.

For sending and receiving between pools that do support large dnodes,
the drr_object send record type is extended with a new field to store
the dnode slot count. This field was repurposed from unused padding
in the structure.

ZIL Replay
----------
The dnode slot count is stored in the uppermost 8 bits of the lr_foid
field. The bits were unused as the object id is currently capped at
48 bits.

Resizing Dnodes
---------------
It should be possible to resize a dnode when it is dirtied if the
current dnodesize dataset property differs from the dnode's size, but
this functionality is not currently implemented. Clearly a dnode can
only grow if there are sufficient contiguous unused slots in the
dnode block, but it should always be possible to shrink a dnode.
Growing dnodes may be useful to reduce fragmentation in a pool with
many spill blocks in use. Shrinking dnodes may be useful to allow
sending a dataset to a pool that doesn't support the large_dnode
feature.

Feature Reference Counting
--------------------------
The reference count for the large_dnode pool feature tracks the
number of datasets that have ever contained a dnode of size larger
than 512 bytes. The first time a large dnode is created in a dataset
the dataset is converted to an extensible dataset. This is a one-way
operation and the only way to decrement the feature count is to
destroy the dataset, even if the dataset no longer contains any large
dnodes. The complexity of reference counting on a per-dnode basis was
too high, so we chose to track it on a per-dataset basis similarly to
the large_block feature.

Signed-off-by: Ned Bass <bass6@llnl.gov>
Signed-off-by: Brian Behlendorf <behlendorf1@llnl.gov>
Closes #3542
2016-06-24 13:13:21 -07:00

462 lines
14 KiB
Groff
Raw Blame History

'\" te
.\" Copyright (c) 2013 by Delphix. All rights reserved.
.\" Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
.\" Copyright (c) 2014, Joyent, Inc. All rights reserved.
.\" The contents of this file are subject to the terms of the Common Development
.\" and Distribution License (the "License"). You may not use this file except
.\" in compliance with the License. You can obtain a copy of the license at
.\" usr/src/OPENSOLARIS.LICENSE or http://www.opensolaris.org/os/licensing.
.\"
.\" See the License for the specific language governing permissions and
.\" limitations under the License. When distributing Covered Code, include this
.\" CDDL HEADER in each file and include the License file at
.\" usr/src/OPENSOLARIS.LICENSE. If applicable, add the following below this
.\" CDDL HEADER, with the fields enclosed by brackets "[]" replaced with your
.\" own identifying information:
.\" Portions Copyright [yyyy] [name of copyright owner]
.TH ZPOOL-FEATURES 5 "Aug 27, 2013"
.SH NAME
zpool\-features \- ZFS pool feature descriptions
.SH DESCRIPTION
.sp
.LP
ZFS pool on\-disk format versions are specified via "features" which replace
the old on\-disk format numbers (the last supported on\-disk format number is
28). To enable a feature on a pool use the \fBupgrade\fR subcommand of the
\fBzpool\fR(8) command, or set the \fBfeature@\fR\fIfeature_name\fR property
to \fBenabled\fR.
.sp
.LP
The pool format does not affect file system version compatibility or the ability
to send file systems between pools.
.sp
.LP
Since most features can be enabled independently of each other the on\-disk
format of the pool is specified by the set of all features marked as
\fBactive\fR on the pool. If the pool was created by another software version
this set may include unsupported features.
.SS "Identifying features"
.sp
.LP
Every feature has a guid of the form \fIcom.example:feature_name\fR. The reverse
DNS name ensures that the feature's guid is unique across all ZFS
implementations. When unsupported features are encountered on a pool they will
be identified by their guids. Refer to the documentation for the ZFS
implementation that created the pool for information about those features.
.sp
.LP
Each supported feature also has a short name. By convention a feature's short
name is the portion of its guid which follows the ':' (e.g.
\fIcom.example:feature_name\fR would have the short name \fIfeature_name\fR),
however a feature's short name may differ across ZFS implementations if
following the convention would result in name conflicts.
.SS "Feature states"
.sp
.LP
Features can be in one of three states:
.sp
.ne 2
.na
\fB\fBactive\fR\fR
.ad
.RS 12n
This feature's on\-disk format changes are in effect on the pool. Support for
this feature is required to import the pool in read\-write mode. If this
feature is not read-only compatible, support is also required to import the pool
in read\-only mode (see "Read\-only compatibility").
.RE
.sp
.ne 2
.na
\fB\fBenabled\fR\fR
.ad
.RS 12n
An administrator has marked this feature as enabled on the pool, but the
feature's on\-disk format changes have not been made yet. The pool can still be
imported by software that does not support this feature, but changes may be made
to the on\-disk format at any time which will move the feature to the
\fBactive\fR state. Some features may support returning to the \fBenabled\fR
state after becoming \fBactive\fR. See feature\-specific documentation for
details.
.RE
.sp
.ne 2
.na
\fBdisabled\fR
.ad
.RS 12n
This feature's on\-disk format changes have not been made and will not be made
unless an administrator moves the feature to the \fBenabled\fR state. Features
cannot be disabled once they have been enabled.
.RE
.sp
.LP
The state of supported features is exposed through pool properties of the form
\fIfeature@short_name\fR.
.SS "Read\-only compatibility"
.sp
.LP
Some features may make on\-disk format changes that do not interfere with other
software's ability to read from the pool. These features are referred to as
"read\-only compatible". If all unsupported features on a pool are read\-only
compatible, the pool can be imported in read\-only mode by setting the
\fBreadonly\fR property during import (see \fBzpool\fR(8) for details on
importing pools).
.SS "Unsupported features"
.sp
.LP
For each unsupported feature enabled on an imported pool a pool property
named \fIunsupported@feature_guid\fR will indicate why the import was allowed
despite the unsupported feature. Possible values for this property are:
.sp
.ne 2
.na
\fB\fBinactive\fR\fR
.ad
.RS 12n
The feature is in the \fBenabled\fR state and therefore the pool's on\-disk
format is still compatible with software that does not support this feature.
.RE
.sp
.ne 2
.na
\fB\fBreadonly\fR\fR
.ad
.RS 12n
The feature is read\-only compatible and the pool has been imported in
read\-only mode.
.RE
.SS "Feature dependencies"
.sp
.LP
Some features depend on other features being enabled in order to function
properly. Enabling a feature will automatically enable any features it
depends on.
.SH FEATURES
.sp
.LP
The following features are supported on this system:
.sp
.ne 2
.na
\fB\fBasync_destroy\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:async_destroy
READ\-ONLY COMPATIBLE yes
DEPENDENCIES none
.TE
Destroying a file system requires traversing all of its data in order to
return its used space to the pool. Without \fBasync_destroy\fR the file system
is not fully removed until all space has been reclaimed. If the destroy
operation is interrupted by a reboot or power outage the next attempt to open
the pool will need to complete the destroy operation synchronously.
When \fBasync_destroy\fR is enabled the file system's data will be reclaimed
by a background process, allowing the destroy operation to complete without
traversing the entire file system. The background process is able to resume
interrupted destroys after the pool has been opened, eliminating the need
to finish interrupted destroys as part of the open operation. The amount
of space remaining to be reclaimed by the background process is available
through the \fBfreeing\fR property.
This feature is only \fBactive\fR while \fBfreeing\fR is non\-zero.
.RE
.sp
.ne 2
.na
\fB\fBempty_bpobj\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:empty_bpobj
READ\-ONLY COMPATIBLE yes
DEPENDENCIES none
.TE
This feature increases the performance of creating and using a large
number of snapshots of a single filesystem or volume, and also reduces
the disk space required.
When there are many snapshots, each snapshot uses many Block Pointer
Objects (bpobj's) to track blocks associated with that snapshot.
However, in common use cases, most of these bpobj's are empty. This
feature allows us to create each bpobj on-demand, thus eliminating the
empty bpobjs.
This feature is \fBactive\fR while there are any filesystems, volumes,
or snapshots which were created after enabling this feature.
.RE
.sp
.ne 2
.na
\fB\fBfilesystem_limits\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.joyent:filesystem_limits
READ\-ONLY COMPATIBLE yes
DEPENDENCIES extensible_dataset
.TE
This feature enables filesystem and snapshot limits. These limits can be used
to control how many filesystems and/or snapshots can be created at the point in
the tree on which the limits are set.
This feature is \fBactive\fR once either of the limit properties has been
set on a dataset. Once activated the feature is never deactivated.
.RE
.sp
.ne 2
.na
\fB\fBlz4_compress\fR\fR
.ad
.RS 4n
.TS
l l .
GUID org.illumos:lz4_compress
READ\-ONLY COMPATIBLE no
DEPENDENCIES none
.TE
\fBlz4\fR is a high-performance real-time compression algorithm that
features significantly faster compression and decompression as well as a
higher compression ratio than the older \fBlzjb\fR compression.
Typically, \fBlz4\fR compression is approximately 50% faster on
compressible data and 200% faster on incompressible data than
\fBlzjb\fR. It is also approximately 80% faster on decompression, while
giving approximately 10% better compression ratio.
When the \fBlz4_compress\fR feature is set to \fBenabled\fR, the
administrator can turn on \fBlz4\fR compression on any dataset on the
pool using the \fBzfs\fR(8) command. Please note that doing so will
immediately activate the \fBlz4_compress\fR feature on the underlying
pool using the \fBzfs\fR(1M) command. Also, all newly written metadata
will be compressed with \fBlz4\fR algorithm. Since this feature is not
read-only compatible, this operation will render the pool unimportable
on systems without support for the \fBlz4_compress\fR feature. Booting
off of \fBlz4\fR-compressed root pools is supported.
This feature becomes \fBactive\fR as soon as it is enabled and will
never return to being \fBenabled\fB.
.RE
.sp
.ne 2
.na
\fB\fBspacemap_histogram\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:spacemap_histogram
READ\-ONLY COMPATIBLE yes
DEPENDENCIES none
.TE
This features allows ZFS to maintain more information about how free space
is organized within the pool. If this feature is \fBenabled\fR, ZFS will
set this feature to \fBactive\fR when a new space map object is created or
an existing space map is upgraded to the new format. Once the feature is
\fBactive\fR, it will remain in that state until the pool is destroyed.
.RE
.sp
.ne 2
.na
\fB\fBextensible_dataset\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:extensible_dataset
READ\-ONLY COMPATIBLE no
DEPENDENCIES none
.TE
This feature allows more flexible use of internal ZFS data structures,
and exists for other features to depend on.
This feature will be \fBactive\fR when the first dependent feature uses it,
and will be returned to the \fBenabled\fR state when all datasets that use
this feature are destroyed.
.RE
.sp
.ne 2
.na
\fB\fBbookmarks\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:bookmarks
READ\-ONLY COMPATIBLE yes
DEPENDENCIES extensible_dataset
.TE
This feature enables use of the \fBzfs bookmark\fR subcommand.
This feature is \fBactive\fR while any bookmarks exist in the pool.
All bookmarks in the pool can be listed by running
\fBzfs list -t bookmark -r \fIpoolname\fR\fR.
.RE
.sp
.ne 2
.na
\fB\fBenabled_txg\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:enabled_txg
READ\-ONLY COMPATIBLE yes
DEPENDENCIES none
.TE
Once this feature is enabled ZFS records the transaction group number
in which new features are enabled. This has no user-visible impact,
but other features may depend on this feature.
This feature becomes \fBactive\fR as soon as it is enabled and will
never return to being \fBenabled\fB.
.RE
.sp
.ne 2
.na
\fB\fBhole_birth\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:hole_birth
READ\-ONLY COMPATIBLE no
DEPENDENCIES enabled_txg
.TE
This feature improves performance of incremental sends ("zfs send -i")
and receives for objects with many holes. The most common case of
hole-filled objects is zvols.
An incremental send stream from snapshot \fBA\fR to snapshot \fBB\fR
contains information about every block that changed between \fBA\fR and
\fBB\fR. Blocks which did not change between those snapshots can be
identified and omitted from the stream using a piece of metadata called
the 'block birth time', but birth times are not recorded for holes (blocks
filled only with zeroes). Since holes created after \fBA\fR cannot be
distinguished from holes created before \fBA\fR, information about every
hole in the entire filesystem or zvol is included in the send stream.
For workloads where holes are rare this is not a problem. However, when
incrementally replicating filesystems or zvols with many holes (for
example a zvol formatted with another filesystem) a lot of time will
be spent sending and receiving unnecessary information about holes that
already exist on the receiving side.
Once the \fBhole_birth\fR feature has been enabled the block birth times
of all new holes will be recorded. Incremental sends between snapshots
created after this feature is enabled will use this new metadata to avoid
sending information about holes that already exist on the receiving side.
This feature becomes \fBactive\fR as soon as it is enabled and will
never return to being \fBenabled\fB.
.RE
.sp
.ne 2
.na
\fB\fBembedded_data\fR\fR
.ad
.RS 4n
.TS
l l .
GUID com.delphix:embedded_data
READ\-ONLY COMPATIBLE no
DEPENDENCIES none
.TE
This feature improves the performance and compression ratio of
highly-compressible blocks. Blocks whose contents can compress to 112 bytes
or smaller can take advantage of this feature.
When this feature is enabled, the contents of highly-compressible blocks are
stored in the block "pointer" itself (a misnomer in this case, as it contains
the compressed data, rather than a pointer to its location on disk). Thus
the space of the block (one sector, typically 512 bytes or 4KB) is saved,
and no additional i/o is needed to read and write the data block.
This feature becomes \fBactive\fR as soon as it is enabled and will
never return to being \fBenabled\fR.
.RE
.sp
.ne 2
.na
\fB\fBlarge_blocks\fR\fR
.ad
.RS 4n
.TS
l l .
GUID org.open-zfs:large_block
READ\-ONLY COMPATIBLE no
DEPENDENCIES extensible_dataset
.TE
The \fBlarge_block\fR feature allows the record size on a dataset to be
set larger than 128KB.
This feature becomes \fBactive\fR once a \fBrecordsize\fR property has been
set larger than 128KB, and will return to being \fBenabled\fR once all
filesystems that have ever had their recordsize larger than 128KB are destroyed.
.RE
.sp
.ne 2
.na
\fB\fBlarge_dnode\fR\fR
.ad
.RS 4n
.TS
l l .
GUID org.zfsonlinux:large_dnode
READ\-ONLY COMPATIBLE no
DEPENDENCIES extensible_dataset
.TE
The \fBlarge_dnode\fR feature allows the size of dnodes in a dataset to be
set larger than 512B.
This feature becomes \fBactive\fR once a dataset contains an object with
a dnode larger than 512B, which occurs as a result of setting the
\fBdnodesize\fR dataset property to a value other than \fBlegacy\fR. The
feature will return to being \fBenabled\fR once all filesystems that
have ever contained a dnode larger than 512B are destroyed. Large dnodes
allow more data to be stored in the bonus buffer, thus potentially
improving performance by avoiding the use of spill blocks.
.RE
.SH "SEE ALSO"
\fBzpool\fR(8)
Reference in New Issue View Git Blame Copy Permalink