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

2 files changed, 44 insertions, 8 deletions

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),