Include scatter_chunk_waste in arc_size

The ARC caches data in scatter ABD's, which are collections of pages,
which are typically 4K.  Therefore, the space used to cache each block
is rounded up to a multiple of 4K.  The ABD subsystem tracks this wasted
memory in the `scatter_chunk_waste` kstat.  However, the ARC's `size` is
not aware of the memory used by this round-up, it only accounts for the
size that it requested from the ABD subsystem.

Therefore, the ARC is effectively using more memory than it is aware of,
due to the `scatter_chunk_waste`.  This impacts observability, e.g.
`arcstat` will show that the ARC is using less memory than it
effectively is.  It also impacts how the ARC responds to memory
pressure.  As the amount of `scatter_chunk_waste` changes, it appears to
the ARC as memory pressure, so it needs to resize `arc_c`.

If the sector size (`1<<ashift`) is the same as the page size (or
larger), there won't be any waste.  If the (compressed) block size is
relatively large compared to the page size, the amount of
`scatter_chunk_waste` will be small, so the problematic effects are
minimal.

However, if using 512B sectors (`ashift=9`), and the (compressed) block
size is small (e.g. `compression=on` with the default `volblocksize=8k`
or a decreased `recordsize`), the amount of `scatter_chunk_waste` can be
very large.  On a production system, with `arc_size` at a constant 50%
of memory, `scatter_chunk_waste` has been been observed to be 10-30% of
memory.

This commit adds `scatter_chunk_waste` to `arc_size`, and adds a new
`waste` field to `arcstat`.  As a result, the ARC's memory usage is more
observable, and `arc_c` does not need to be adjusted as frequently.

Reviewed-by: Pavel Zakharov <[email protected]>
Reviewed-by: Brian Behlendorf <[email protected]>
Reviewed-by: George Wilson <[email protected]>
Reviewed-by: Ryan Moeller <[email protected]>
Signed-off-by: Matthew Ahrens <[email protected]>
Closes #10701 

1 files changed, 20 insertions, 2 deletions

diff --git a/module/zfs/arc.c b/module/zfs/arc.c
index 06c2d5fac..f63f92b86 100644
--- a/module/zfs/arc.c
+++ b/module/zfs/arc.c
@@ -583,6 +583,7 @@ arc_stats_t arc_stats = {
 	{ "arc_sys_free",		KSTAT_DATA_UINT64 },
 	{ "arc_raw_size",		KSTAT_DATA_UINT64 },
 	{ "cached_only_in_progress",	KSTAT_DATA_UINT64 },
+	{ "abd_chunk_waste_size",	KSTAT_DATA_UINT64 },
 };
 
 #define	ARCSTAT_MAX(stat, val) {					\
@@ -685,6 +686,7 @@ aggsum_t astat_dnode_size;
 aggsum_t astat_bonus_size;
 aggsum_t astat_hdr_size;
 aggsum_t astat_l2_hdr_size;
+aggsum_t astat_abd_chunk_waste_size;
 
 hrtime_t arc_growtime;
 list_t arc_prune_list;
@@ -2611,9 +2613,18 @@ arc_space_consume(uint64_t space, arc_space_type_t type)
 	case ARC_SPACE_L2HDRS:
 		aggsum_add(&astat_l2_hdr_size, space);
 		break;
+	case ARC_SPACE_ABD_CHUNK_WASTE:
+		/*
+		 * Note: this includes space wasted by all scatter ABD's, not
+		 * just those allocated by the ARC.  But the vast majority of
+		 * scatter ABD's come from the ARC, because other users are
+		 * very short-lived.
+		 */
+		aggsum_add(&astat_abd_chunk_waste_size, space);
+		break;
 	}
 
-	if (type != ARC_SPACE_DATA)
+	if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE)
 		aggsum_add(&arc_meta_used, space);
 
 	aggsum_add(&arc_size, space);
@@ -2648,9 +2659,12 @@ arc_space_return(uint64_t space, arc_space_type_t type)
 	case ARC_SPACE_L2HDRS:
 		aggsum_add(&astat_l2_hdr_size, -space);
 		break;
+	case ARC_SPACE_ABD_CHUNK_WASTE:
+		aggsum_add(&astat_abd_chunk_waste_size, -space);
+		break;
 	}
 
-	if (type != ARC_SPACE_DATA) {
+	if (type != ARC_SPACE_DATA && type != ARC_SPACE_ABD_CHUNK_WASTE) {
 		ASSERT(aggsum_compare(&arc_meta_used, space) >= 0);
 		/*
 		 * We use the upper bound here rather than the precise value
@@ -7090,6 +7104,8 @@ arc_kstat_update(kstat_t *ksp, int rw)
 		ARCSTAT(arcstat_dbuf_size) = aggsum_value(&astat_dbuf_size);
 		ARCSTAT(arcstat_dnode_size) = aggsum_value(&astat_dnode_size);
 		ARCSTAT(arcstat_bonus_size) = aggsum_value(&astat_bonus_size);
+		ARCSTAT(arcstat_abd_chunk_waste_size) =
+		    aggsum_value(&astat_abd_chunk_waste_size);
 
 		as->arcstat_memory_all_bytes.value.ui64 =
 		    arc_all_memory();
@@ -7329,6 +7345,7 @@ arc_state_init(void)
 	aggsum_init(&astat_bonus_size, 0);
 	aggsum_init(&astat_dnode_size, 0);
 	aggsum_init(&astat_dbuf_size, 0);
+	aggsum_init(&astat_abd_chunk_waste_size, 0);
 
 	arc_anon->arcs_state = ARC_STATE_ANON;
 	arc_mru->arcs_state = ARC_STATE_MRU;
@@ -7381,6 +7398,7 @@ arc_state_fini(void)
 	aggsum_fini(&astat_bonus_size);
 	aggsum_fini(&astat_dnode_size);
 	aggsum_fini(&astat_dbuf_size);
+	aggsum_fini(&astat_abd_chunk_waste_size);
 }
 
 uint64_t