Illumos 3835 zfs need not store 2 copies of all metadata

Reviewed by: George Wilson <[email protected]>
Reviewed by: Adam Leventhal <[email protected]>
Reviewed by: Dan McDonald <[email protected]>
Approved by: Richard Lowe <[email protected]>

Description from Matt Ahrens's bug report at Delphix:

    Add a new zfs property, "redundant_metadata" which can have values
    "all" or "most".  The default will be "all", which is the current
    behavior.  Setting to "most" will cause us to only store 1 copy of
    level-1 indirect blocks of user data files.

Additional notes:

    The new man page section for this property states

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

    and:

        "When set to most, ZFS stores an extra copy of most types of
         metadata. This can improve performance of random writes,
         because less metadata must be written."

    The current implementation is as described above in Matt's blog.
    It is controlled by a new global integer
    "zfs_redundant_metadata_most_ditto_level", currently initialized
    to 2. When "redundant_metadata" is set to "most", only indirect
    blocks of the specified level and higher will have additional ditto
    blocks created.

Ported by: Tim Chase <[email protected]>
Signed-off-by: Brian Behlendorf <[email protected]>
Closes #2542

7 files changed, 108 insertions, 23 deletions

diff --git a/include/sys/dmu.h b/include/sys/dmu.h
index 8e25acf85..ad12ecafa 100644
--- a/include/sys/dmu.h
+++ b/include/sys/dmu.h
@@ -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 2011 Nexenta Systems, Inc. All rights reserved.
  * Copyright (c) 2012, Joyent, Inc. All rights reserved.
  */
@@ -739,8 +739,8 @@ 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_lookup(objset_t *os, const char *name, uint64_t *val);
diff --git a/include/sys/dmu_objset.h b/include/sys/dmu_objset.h
index edf362f7f..cbf0394e6 100644
--- a/include/sys/dmu_objset.h
+++ b/include/sys/dmu_objset.h
@@ -21,7 +21,7 @@
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
  * Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
- * Copyright (c) 2012 by Delphix. All rights reserved.
+ * Copyright (c) 2012, 2014 by Delphix. All rights reserved.
  */
 
 /* Portions Copyright 2010 Robert Milkowski */
@@ -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 */
diff --git a/include/sys/fs/zfs.h b/include/sys/fs/zfs.h
index 53f2718ab..5371e12ef 100644
--- a/include/sys/fs/zfs.h
+++ b/include/sys/fs/zfs.h
@@ -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.
  */
@@ -149,6 +149,7 @@ typedef enum {
 	ZFS_PROP_SELINUX_DEFCONTEXT,
 	ZFS_PROP_SELINUX_ROOTCONTEXT,
 	ZFS_PROP_RELATIME,
+	ZFS_PROP_REDUNDANT_METADATA,
 	ZFS_NUM_PROPS
 } zfs_prop_t;
 
@@ -349,6 +350,11 @@ typedef enum {
 	ZFS_XATTR_SA = 2
 } zfs_xattr_type_t;
 
+typedef enum {
+	ZFS_REDUNDANT_METADATA_ALL,
+	ZFS_REDUNDANT_METADATA_MOST
+} zfs_redundant_metadata_type_t;
+
 /*
  * On-disk version number.
  */
diff --git a/man/man8/zfs.8 b/man/man8/zfs.8
index 09d63ba5a..5683206ee 100644
--- a/man/man8/zfs.8
+++ b/man/man8/zfs.8
@@ -22,7 +22,7 @@
 .\"
 .\" Copyright (c) 2009 Sun Microsystems, Inc. All Rights Reserved.
 .\" Copyright 2011 Joshua M. Clulow <[email protected]>
-.\" Copyright (c) 2013 by Delphix. All rights reserved.
+.\" Copyright (c) 2014 by Delphix. All rights reserved.
 .\" Copyright (c) 2012, Joyent, Inc. All rights reserved.
 .\" Copyright 2012 Nexenta Systems, Inc. All Rights Reserved.
 .\" Copyright (c) 2013 by Saso Kiselkov. All rights reserved.
@@ -1014,6 +1014,37 @@ This property can also be referred to by its shortened column name, \fBrecsize\f
 .ne 2
 .mk
 .na
+\fB\fBredundant_metadata\fR=\fBall\fR | \fBmost\fR\fR
+.ad
+.sp .6
+.RS 4n
+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 (e.g. by mirroring or RAID-Z),
+and is in addition to an extra copy specified by the \fBcopies\fR
+property (up to a total of 3 copies).  For example if the pool is
+mirrored, \fBcopies\fR=2, and \fBredundant_metadata\fR=most, then ZFS
+stores 6 copies of most metadata, and 4 copies of data and some
+metadata.
+.sp
+When set to \fBall\fR, 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 \fBrecordsize\fR bytes long) can be lost.
+.sp
+When set to \fBmost\fR, 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 (of
+\fBrecordsize\fR 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.
+.sp
+The default value is \fBall\fR.
+.RE
+
+.sp
+.ne 2
+.na
 \fB\fBrefquota\fR=\fIsize\fR | \fBnone\fR\fR
 .ad
 .sp .6
diff --git a/module/zcommon/zfs_prop.c b/module/zcommon/zfs_prop.c
index 92cfa282b..77491cdb7 100644
--- a/module/zcommon/zfs_prop.c
+++ b/module/zcommon/zfs_prop.c
@@ -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.
  */
 
@@ -210,7 +210,18 @@ 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",
diff --git a/module/zfs/dmu.c b/module/zfs/dmu.c
index 103eec1d5..d8e5739f4 100644
--- a/module/zfs/dmu.c
+++ b/module/zfs/dmu.c
@@ -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.
  */
 
@@ -1688,6 +1688,12 @@ dmu_object_set_compress(objset_t *os, uint64_t object, uint8_t compress,
 
 int zfs_mdcomp_disable = 0;
 
+/*
+ * 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)
 {
@@ -1727,6 +1733,13 @@ 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);
 
@@ -1774,7 +1787,7 @@ 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;
diff --git a/module/zfs/dmu_objset.c b/module/zfs/dmu_objset.c
index fc7c80365..b82783098 100644
--- a/module/zfs/dmu_objset.c
+++ b/module/zfs/dmu_objset.c
@@ -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.
  */
 
@@ -114,13 +114,13 @@ 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)
 }
 
 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)
 {
 	objset_t *os = arg;
@@ -363,6 +377,12 @@ 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,
@@ -376,9 +396,9 @@ 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;
 	}
@@ -623,6 +643,9 @@ 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),