Pool allocation classes

Allocation Classes add the ability to have allocation classes in a
pool that are dedicated to serving specific block categories, such
as DDT data, metadata, and small file blocks. A pool can opt-in to
this feature by adding a 'special' or 'dedup' top-level VDEV.

Reviewed by: Pavel Zakharov <[email protected]>
Reviewed-by: Richard Laager <[email protected]>
Reviewed-by: Alek Pinchuk <[email protected]>
Reviewed-by: Håkan Johansson <[email protected]>
Reviewed-by: Andreas Dilger <[email protected]>
Reviewed-by: DHE <[email protected]>
Reviewed-by: Richard Elling <[email protected]>
Reviewed-by: Gregor Kopka <[email protected]>
Reviewed-by: Kash Pande <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: Matthew Ahrens <[email protected]>
Signed-off-by: Don Brady <[email protected]>
Closes #5182 

1 files changed, 105 insertions, 23 deletions

diff --git a/cmd/zdb/zdb.c b/cmd/zdb/zdb.c
index 38ffbbece..21113da2f 100644
--- a/cmd/zdb/zdb.c
+++ b/cmd/zdb/zdb.c
@@ -21,7 +21,7 @@
 
 /*
  * Copyright (c) 2005, 2010, Oracle and/or its affiliates. All rights reserved.
- * Copyright (c) 2011, 2017 by Delphix. All rights reserved.
+ * Copyright (c) 2011, 2018 by Delphix. All rights reserved.
  * Copyright (c) 2014 Integros [integros.com]
  * Copyright 2016 Nexenta Systems, Inc.
  * Copyright (c) 2017, 2018 Lawrence Livermore National Security, LLC.
@@ -932,13 +932,23 @@ dump_metaslab(metaslab_t *msp)
 static void
 print_vdev_metaslab_header(vdev_t *vd)
 {
-	(void) printf("\tvdev %10llu\n\t%-10s%5llu   %-19s   %-15s   %-10s\n",
-	    (u_longlong_t)vd->vdev_id,
+	vdev_alloc_bias_t alloc_bias = vd->vdev_alloc_bias;
+	const char *bias_str;
+
+	bias_str = (alloc_bias == VDEV_BIAS_LOG || vd->vdev_islog) ?
+	    VDEV_ALLOC_BIAS_LOG :
+	    (alloc_bias == VDEV_BIAS_SPECIAL) ? VDEV_ALLOC_BIAS_SPECIAL :
+	    (alloc_bias == VDEV_BIAS_DEDUP) ? VDEV_ALLOC_BIAS_DEDUP :
+	    vd->vdev_islog ? "log" : "";
+
+	(void) printf("\tvdev %10llu   %s\n"
+	    "\t%-10s%5llu   %-19s   %-15s   %-12s\n",
+	    (u_longlong_t)vd->vdev_id, bias_str,
 	    "metaslabs", (u_longlong_t)vd->vdev_ms_count,
 	    "offset", "spacemap", "free");
-	(void) printf("\t%15s   %19s   %15s   %10s\n",
+	(void) printf("\t%15s   %19s   %15s   %12s\n",
 	    "---------------", "-------------------",
-	    "---------------", "-------------");
+	    "---------------", "------------");
 }
 
 static void
@@ -954,7 +964,7 @@ dump_metaslab_groups(spa_t *spa)
 		vdev_t *tvd = rvd->vdev_child[c];
 		metaslab_group_t *mg = tvd->vdev_mg;
 
-		if (mg->mg_class != mc)
+		if (mg == NULL || mg->mg_class != mc)
 			continue;
 
 		metaslab_group_histogram_verify(mg);
@@ -3158,6 +3168,7 @@ typedef struct zdb_blkstats {
 	uint64_t zb_count;
 	uint64_t zb_gangs;
 	uint64_t zb_ditto_samevdev;
+	uint64_t zb_ditto_same_ms;
 	uint64_t zb_psize_histogram[PSIZE_HISTO_SIZE];
 } zdb_blkstats_t;
 
@@ -3197,6 +3208,16 @@ typedef struct zdb_cb {
 	uint32_t	**zcb_vd_obsolete_counts;
 } zdb_cb_t;
 
+/* test if two DVA offsets from same vdev are within the same metaslab */
+static boolean_t
+same_metaslab(spa_t *spa, uint64_t vdev, uint64_t off1, uint64_t off2)
+{
+	vdev_t *vd = vdev_lookup_top(spa, vdev);
+	uint64_t ms_shift = vd->vdev_ms_shift;
+
+	return ((off1 >> ms_shift) == (off2 >> ms_shift));
+}
+
 static void
 zdb_count_block(zdb_cb_t *zcb, zilog_t *zilog, const blkptr_t *bp,
     dmu_object_type_t type)
@@ -3209,6 +3230,8 @@ zdb_count_block(zdb_cb_t *zcb, zilog_t *zilog, const blkptr_t *bp,
 	if (zilog && zil_bp_tree_add(zilog, bp) != 0)
 		return;
 
+	spa_config_enter(zcb->zcb_spa, SCL_CONFIG, FTAG, RW_READER);
+
 	for (i = 0; i < 4; i++) {
 		int l = (i < 2) ? BP_GET_LEVEL(bp) : ZB_TOTAL;
 		int t = (i & 1) ? type : ZDB_OT_TOTAL;
@@ -3234,8 +3257,15 @@ zdb_count_block(zdb_cb_t *zcb, zilog_t *zilog, const blkptr_t *bp,
 		switch (BP_GET_NDVAS(bp)) {
 		case 2:
 			if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
-			    DVA_GET_VDEV(&bp->blk_dva[1]))
+			    DVA_GET_VDEV(&bp->blk_dva[1])) {
 				zb->zb_ditto_samevdev++;
+
+				if (same_metaslab(zcb->zcb_spa,
+				    DVA_GET_VDEV(&bp->blk_dva[0]),
+				    DVA_GET_OFFSET(&bp->blk_dva[0]),
+				    DVA_GET_OFFSET(&bp->blk_dva[1])))
+					zb->zb_ditto_same_ms++;
+			}
 			break;
 		case 3:
 			equal = (DVA_GET_VDEV(&bp->blk_dva[0]) ==
@@ -3244,13 +3274,37 @@ zdb_count_block(zdb_cb_t *zcb, zilog_t *zilog, const blkptr_t *bp,
 			    DVA_GET_VDEV(&bp->blk_dva[2])) +
 			    (DVA_GET_VDEV(&bp->blk_dva[1]) ==
 			    DVA_GET_VDEV(&bp->blk_dva[2]));
-			if (equal != 0)
+			if (equal != 0) {
 				zb->zb_ditto_samevdev++;
+
+				if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
+				    DVA_GET_VDEV(&bp->blk_dva[1]) &&
+				    same_metaslab(zcb->zcb_spa,
+				    DVA_GET_VDEV(&bp->blk_dva[0]),
+				    DVA_GET_OFFSET(&bp->blk_dva[0]),
+				    DVA_GET_OFFSET(&bp->blk_dva[1])))
+					zb->zb_ditto_same_ms++;
+				else if (DVA_GET_VDEV(&bp->blk_dva[0]) ==
+				    DVA_GET_VDEV(&bp->blk_dva[2]) &&
+				    same_metaslab(zcb->zcb_spa,
+				    DVA_GET_VDEV(&bp->blk_dva[0]),
+				    DVA_GET_OFFSET(&bp->blk_dva[0]),
+				    DVA_GET_OFFSET(&bp->blk_dva[2])))
+					zb->zb_ditto_same_ms++;
+				else if (DVA_GET_VDEV(&bp->blk_dva[1]) ==
+				    DVA_GET_VDEV(&bp->blk_dva[2]) &&
+				    same_metaslab(zcb->zcb_spa,
+				    DVA_GET_VDEV(&bp->blk_dva[1]),
+				    DVA_GET_OFFSET(&bp->blk_dva[1]),
+				    DVA_GET_OFFSET(&bp->blk_dva[2])))
+					zb->zb_ditto_same_ms++;
+			}
 			break;
 		}
-
 	}
 
+	spa_config_exit(zcb->zcb_spa, SCL_CONFIG, FTAG);
+
 	if (BP_IS_EMBEDDED(bp)) {
 		zcb->zcb_embedded_blocks[BPE_GET_ETYPE(bp)]++;
 		zcb->zcb_embedded_histogram[BPE_GET_ETYPE(bp)]
@@ -4103,6 +4157,8 @@ dump_block_stats(spa_t *spa)
 		flags |= TRAVERSE_PREFETCH_DATA;
 
 	zcb.zcb_totalasize = metaslab_class_get_alloc(spa_normal_class(spa));
+	zcb.zcb_totalasize += metaslab_class_get_alloc(spa_special_class(spa));
+	zcb.zcb_totalasize += metaslab_class_get_alloc(spa_dedup_class(spa));
 	zcb.zcb_start = zcb.zcb_lastprint = gethrtime();
 	err = traverse_pool(spa, 0, flags, zdb_blkptr_cb, &zcb);
 
@@ -4147,7 +4203,10 @@ dump_block_stats(spa_t *spa)
 	norm_alloc = metaslab_class_get_alloc(spa_normal_class(spa));
 	norm_space = metaslab_class_get_space(spa_normal_class(spa));
 
-	total_alloc = norm_alloc + metaslab_class_get_alloc(spa_log_class(spa));
+	total_alloc = norm_alloc +
+	    metaslab_class_get_alloc(spa_log_class(spa)) +
+	    metaslab_class_get_alloc(spa_special_class(spa)) +
+	    metaslab_class_get_alloc(spa_dedup_class(spa));
 	total_found = tzb->zb_asize - zcb.zcb_dedup_asize +
 	    zcb.zcb_removing_size + zcb.zcb_checkpoint_size;
 
@@ -4169,31 +4228,50 @@ dump_block_stats(spa_t *spa)
 		return (2);
 
 	(void) printf("\n");
-	(void) printf("\tbp count:      %10llu\n",
+	(void) printf("\t%-16s %14llu\n", "bp count:",
 	    (u_longlong_t)tzb->zb_count);
-	(void) printf("\tganged count:  %10llu\n",
+	(void) printf("\t%-16s %14llu\n", "ganged count:",
 	    (longlong_t)tzb->zb_gangs);
-	(void) printf("\tbp logical:    %10llu      avg: %6llu\n",
+	(void) printf("\t%-16s %14llu      avg: %6llu\n", "bp logical:",
 	    (u_longlong_t)tzb->zb_lsize,
 	    (u_longlong_t)(tzb->zb_lsize / tzb->zb_count));
-	(void) printf("\tbp physical:   %10llu      avg:"
-	    " %6llu     compression: %6.2f\n",
-	    (u_longlong_t)tzb->zb_psize,
+	(void) printf("\t%-16s %14llu      avg: %6llu     compression: %6.2f\n",
+	    "bp physical:", (u_longlong_t)tzb->zb_psize,
 	    (u_longlong_t)(tzb->zb_psize / tzb->zb_count),
 	    (double)tzb->zb_lsize / tzb->zb_psize);
-	(void) printf("\tbp allocated:  %10llu      avg:"
-	    " %6llu     compression: %6.2f\n",
-	    (u_longlong_t)tzb->zb_asize,
+	(void) printf("\t%-16s %14llu      avg: %6llu     compression: %6.2f\n",
+	    "bp allocated:", (u_longlong_t)tzb->zb_asize,
 	    (u_longlong_t)(tzb->zb_asize / tzb->zb_count),
 	    (double)tzb->zb_lsize / tzb->zb_asize);
-	(void) printf("\tbp deduped:    %10llu    ref>1:"
-	    " %6llu   deduplication: %6.2f\n",
-	    (u_longlong_t)zcb.zcb_dedup_asize,
+	(void) printf("\t%-16s %14llu    ref>1: %6llu   deduplication: %6.2f\n",
+	    "bp deduped:", (u_longlong_t)zcb.zcb_dedup_asize,
 	    (u_longlong_t)zcb.zcb_dedup_blocks,
 	    (double)zcb.zcb_dedup_asize / tzb->zb_asize + 1.0);
-	(void) printf("\tSPA allocated: %10llu     used: %5.2f%%\n",
+	(void) printf("\t%-16s %14llu     used: %5.2f%%\n", "Normal class:",
 	    (u_longlong_t)norm_alloc, 100.0 * norm_alloc / norm_space);
 
+	if (spa_special_class(spa)->mc_rotor != NULL) {
+		uint64_t alloc = metaslab_class_get_alloc(
+		    spa_special_class(spa));
+		uint64_t space = metaslab_class_get_space(
+		    spa_special_class(spa));
+
+		(void) printf("\t%-16s %14llu     used: %5.2f%%\n",
+		    "Special class", (u_longlong_t)alloc,
+		    100.0 * alloc / space);
+	}
+
+	if (spa_dedup_class(spa)->mc_rotor != NULL) {
+		uint64_t alloc = metaslab_class_get_alloc(
+		    spa_dedup_class(spa));
+		uint64_t space = metaslab_class_get_space(
+		    spa_dedup_class(spa));
+
+		(void) printf("\t%-16s %14llu     used: %5.2f%%\n",
+		    "Dedup class", (u_longlong_t)alloc,
+		    100.0 * alloc / space);
+	}
+
 	for (i = 0; i < NUM_BP_EMBEDDED_TYPES; i++) {
 		if (zcb.zcb_embedded_blocks[i] == 0)
 			continue;
@@ -4215,6 +4293,10 @@ dump_block_stats(spa_t *spa)
 		(void) printf("\tDittoed blocks on same vdev: %llu\n",
 		    (longlong_t)tzb->zb_ditto_samevdev);
 	}
+	if (tzb->zb_ditto_same_ms != 0) {
+		(void) printf("\tDittoed blocks in same metaslab: %llu\n",
+		    (longlong_t)tzb->zb_ditto_same_ms);
+	}
 
 	for (uint64_t v = 0; v < spa->spa_root_vdev->vdev_children; v++) {
 		vdev_t *vd = spa->spa_root_vdev->vdev_child[v];