MFV r266766:

Add a new zfs property, "redundant_metadata" which can have values "all" or
"most".  The default will be "all", which is the current behavior.  When set
to all, ZFS stores an extra copy of all metadata.  If a single on-disk block
is corrupt, at worst a single block of user data (which is recordsize bytes
long) can be lost.

Setting to "most" will cause us to only store 1 copy of level-1 indirect
blocks of user data files.  This can improve performance of random writes,
because less metadata has to be written.  In practice,  at worst about
100 blocks (of recordsize bytes each) of user data can be lost if a single
on-disk block is corrupt.

The exact behavior of which metadata blocks are stored redundantly may change
in future releases.

Illumos issue: 3835 zfs need not store 2 copies of all metadata

MFC after:	2 weeks
This commit is contained in:
delphij 2014-05-27 19:46:11 +00:00
parent 0c27a75607
commit 78ba176089
7 changed files with 125 additions and 24 deletions

View File

@ -18,7 +18,7 @@
.\" information: Portions Copyright [yyyy] [name of copyright owner]
.\"
.\" Copyright (c) 2010, Sun Microsystems, Inc. All Rights Reserved.
.\" Copyright (c) 2013 by Delphix. All rights reserved.
.\" Copyright (c) 2014 by Delphix. All rights reserved.
.\" Copyright (c) 2011, Pawel Jakub Dawidek <pjd@FreeBSD.org>
.\" Copyright (c) 2012, Glen Barber <gjb@FreeBSD.org>
.\" Copyright (c) 2012, Bryan Drewery <bdrewery@FreeBSD.org>
@ -30,7 +30,7 @@
.\"
.\" $FreeBSD$
.\"
.Dd April 23, 2014
.Dd May 27, 2014
.Dt ZFS 8
.Os
.Sh NAME
@ -1193,6 +1193,53 @@ affects only files created afterward; existing files are unaffected.
.Pp
This property can also be referred to by its shortened column name,
.Sy recsize .
.It Sy redundant_metadata Ns = Ns Cm all | most
Controls what types of metadata are stored redundantly.
ZFS stores an extra copy of metadata, so that if a single block is corrupted,
the amount of user data lost is limited.
This extra copy is in addition to any redundancy provided at the pool level
.Pq e.g. by mirroring or RAID-Z ,
and is in addition to an extra copy specified by the
.Sy copies
property
.Pq up to a total of 3 copies .
For example if the pool is mirrored,
.Cm copies Ns = Ns Ar 2 ,
and
.Cm redundant_metadata Ns = Ns Ar most ,
then ZFS
stores 6 copies of most metadata, and 4 copies of data and some
metadata.
.Pp
When set to
.Cm all ,
ZFS stores an extra copy of all metadata.
If a
single on-disk block is corrupt, at worst a single block of user data
.Po which is
.Cm recordsize
bytes long
can be lost.
.Pc
.Pp
When set to
.Cm most ,
ZFS stores an extra copy of most types of
metadata.
This can improve performance of random writes, because less
metadata must be written.
In practice, at worst about 100 blocks
.Po of
.Cm recordsize
bytes each
.Pc
of user data can be lost if a single
on-disk block is corrupt.
The exact behavior of which metadata blocks
are stored redundantly may change in future releases.
.Pp
The default value is
.Cm all .
.It Sy refquota Ns = Ns Ar size | Cm none
Limits the amount of space a dataset can consume. This property enforces a hard
limit on the amount of space used. This hard limit does not include space used

View File

@ -20,7 +20,7 @@
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright (c) 2011, 2014 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
*/
@ -206,7 +206,18 @@ zfs_prop_init(void)
{ NULL }
};
static zprop_index_t redundant_metadata_table[] = {
{ "all", ZFS_REDUNDANT_METADATA_ALL },
{ "most", ZFS_REDUNDANT_METADATA_MOST },
{ NULL }
};
/* inherit index properties */
zprop_register_index(ZFS_PROP_REDUNDANT_METADATA, "redundant_metadata",
ZFS_REDUNDANT_METADATA_ALL,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"all | most", "REDUND_MD",
redundant_metadata_table);
zprop_register_index(ZFS_PROP_SYNC, "sync", ZFS_SYNC_STANDARD,
PROP_INHERIT, ZFS_TYPE_FILESYSTEM | ZFS_TYPE_VOLUME,
"standard | always | disabled", "SYNC",

View File

@ -20,7 +20,7 @@
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
*/
/* Copyright (c) 2013 by Saso Kiselkov. All rights reserved. */
/* Copyright (c) 2013, Joyent, Inc. All rights reserved. */
@ -1630,6 +1630,12 @@ TUNABLE_INT("vfs.zfs.mdcomp_disable", &zfs_mdcomp_disable);
SYSCTL_INT(_vfs_zfs, OID_AUTO, mdcomp_disable, CTLFLAG_RW,
&zfs_mdcomp_disable, 0, "Disable metadata compression");
/*
* When the "redundant_metadata" property is set to "most", only indirect
* blocks of this level and higher will have an additional ditto block.
*/
int zfs_redundant_metadata_most_ditto_level = 2;
void
dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp)
{
@ -1669,6 +1675,13 @@ dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp)
if (zio_checksum_table[checksum].ci_correctable < 1 ||
zio_checksum_table[checksum].ci_eck)
checksum = ZIO_CHECKSUM_FLETCHER_4;
if (os->os_redundant_metadata == ZFS_REDUNDANT_METADATA_ALL ||
(os->os_redundant_metadata ==
ZFS_REDUNDANT_METADATA_MOST &&
(level >= zfs_redundant_metadata_most_ditto_level ||
DMU_OT_IS_METADATA(type) || (wp & WP_SPILL))))
copies++;
} else if (wp & WP_NOFILL) {
ASSERT(level == 0);
@ -1716,7 +1729,7 @@ dmu_write_policy(objset_t *os, dnode_t *dn, int level, int wp, zio_prop_t *zp)
zp->zp_compress = compress;
zp->zp_type = (wp & WP_SPILL) ? dn->dn_bonustype : type;
zp->zp_level = level;
zp->zp_copies = MIN(copies + ismd, spa_max_replication(os->os_spa));
zp->zp_copies = MIN(copies, spa_max_replication(os->os_spa));
zp->zp_dedup = dedup;
zp->zp_dedup_verify = dedup && dedup_verify;
zp->zp_nopwrite = nopwrite;

View File

@ -20,7 +20,7 @@
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
*/
@ -115,13 +115,13 @@ dmu_objset_id(objset_t *os)
return (ds ? ds->ds_object : 0);
}
uint64_t
zfs_sync_type_t
dmu_objset_syncprop(objset_t *os)
{
return (os->os_sync);
}
uint64_t
zfs_logbias_op_t
dmu_objset_logbias(objset_t *os)
{
return (os->os_logbias);
@ -229,6 +229,20 @@ sync_changed_cb(void *arg, uint64_t newval)
zil_set_sync(os->os_zil, newval);
}
static void
redundant_metadata_changed_cb(void *arg, uint64_t newval)
{
objset_t *os = arg;
/*
* Inheritance and range checking should have been done by now.
*/
ASSERT(newval == ZFS_REDUNDANT_METADATA_ALL ||
newval == ZFS_REDUNDANT_METADATA_MOST);
os->os_redundant_metadata = newval;
}
static void
logbias_changed_cb(void *arg, uint64_t newval)
{
@ -364,6 +378,12 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
zfs_prop_to_name(ZFS_PROP_SYNC),
sync_changed_cb, os);
}
if (err == 0) {
err = dsl_prop_register(ds,
zfs_prop_to_name(
ZFS_PROP_REDUNDANT_METADATA),
redundant_metadata_changed_cb, os);
}
}
if (err != 0) {
VERIFY(arc_buf_remove_ref(os->os_phys_buf,
@ -377,9 +397,9 @@ dmu_objset_open_impl(spa_t *spa, dsl_dataset_t *ds, blkptr_t *bp,
os->os_compress = ZIO_COMPRESS_LZJB;
os->os_copies = spa_max_replication(spa);
os->os_dedup_checksum = ZIO_CHECKSUM_OFF;
os->os_dedup_verify = 0;
os->os_logbias = 0;
os->os_sync = 0;
os->os_dedup_verify = B_FALSE;
os->os_logbias = ZFS_LOGBIAS_LATENCY;
os->os_sync = ZFS_SYNC_STANDARD;
os->os_primary_cache = ZFS_CACHE_ALL;
os->os_secondary_cache = ZFS_CACHE_ALL;
}
@ -622,6 +642,9 @@ dmu_objset_evict(objset_t *os)
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_SYNC),
sync_changed_cb, os));
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_REDUNDANT_METADATA),
redundant_metadata_changed_cb, os));
}
VERIFY0(dsl_prop_unregister(ds,
zfs_prop_to_name(ZFS_PROP_PRIMARYCACHE),

View File

@ -21,7 +21,7 @@
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright (c) 2011, 2014 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2012, Joyent, Inc. All rights reserved.
* Copyright 2013 DEY Storage Systems, Inc.
@ -746,8 +746,8 @@ extern struct dsl_dataset *dmu_objset_ds(objset_t *os);
extern void dmu_objset_name(objset_t *os, char *buf);
extern dmu_objset_type_t dmu_objset_type(objset_t *os);
extern uint64_t dmu_objset_id(objset_t *os);
extern uint64_t dmu_objset_syncprop(objset_t *os);
extern uint64_t dmu_objset_logbias(objset_t *os);
extern zfs_sync_type_t dmu_objset_syncprop(objset_t *os);
extern zfs_logbias_op_t dmu_objset_logbias(objset_t *os);
extern int dmu_snapshot_list_next(objset_t *os, int namelen, char *name,
uint64_t *id, uint64_t *offp, boolean_t *case_conflict);
extern int dmu_snapshot_realname(objset_t *os, char *name, char *real,

View File

@ -20,7 +20,7 @@
*/
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2012 by Delphix. All rights reserved.
* Copyright (c) 2012, 2014 by Delphix. All rights reserved.
* Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
*/
@ -85,15 +85,16 @@ struct objset {
zilog_t *os_zil;
/* can change, under dsl_dir's locks: */
uint8_t os_checksum;
uint8_t os_compress;
enum zio_checksum os_checksum;
enum zio_compress os_compress;
uint8_t os_copies;
uint8_t os_dedup_checksum;
uint8_t os_dedup_verify;
uint8_t os_logbias;
uint8_t os_primary_cache;
uint8_t os_secondary_cache;
uint8_t os_sync;
enum zio_checksum os_dedup_checksum;
boolean_t os_dedup_verify;
zfs_logbias_op_t os_logbias;
zfs_cache_type_t os_primary_cache;
zfs_cache_type_t os_secondary_cache;
zfs_sync_type_t os_sync;
zfs_redundant_metadata_type_t os_redundant_metadata;
/* no lock needed: */
struct dmu_tx *os_synctx; /* XXX sketchy */

View File

@ -21,7 +21,7 @@
/*
* Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
* Copyright (c) 2013 by Delphix. All rights reserved.
* Copyright (c) 2011, 2014 by Delphix. All rights reserved.
* Copyright 2011 Nexenta Systems, Inc. All rights reserved.
* Copyright (c) 2013, Joyent, Inc. All rights reserved.
* Copyright (c) 2012, Martin Matuska <mm@FreeBSD.org>. All rights reserved.
@ -150,6 +150,7 @@ typedef enum {
ZFS_PROP_SNAPSHOT_LIMIT,
ZFS_PROP_FILESYSTEM_COUNT,
ZFS_PROP_SNAPSHOT_COUNT,
ZFS_PROP_REDUNDANT_METADATA,
ZFS_NUM_PROPS
} zfs_prop_t;
@ -349,6 +350,11 @@ typedef enum {
ZFS_VOLMODE_NONE = 3
} zfs_volmode_t;
typedef enum {
ZFS_REDUNDANT_METADATA_ALL,
ZFS_REDUNDANT_METADATA_MOST
} zfs_redundant_metadata_type_t;
/*
* On-disk version number.
*/