ddt: dedup log

Adds a log/journal to dedup. At the end of txg, instead of writing the
entry directly to the ZAP, instead its adding to an in-memory tree and
appended to an on-disk object. The on-disk object is only read at
import, to reload the in-memory tree.

Lookups first go the the log tree before going to the ZAP, so
recently-used entries will remain close by in memory. This vastly
reduces overhead from dedup IO, as it will not have to do so many
read/update/write cycles on ZAP leaf nodes.

A flushing facility is added at end of txg, to push logged entries out
to the ZAP. There's actually two separate "logs" (in-memory tree and
on-disk object), one active (recieving updated entries) and one flushing
(writing out to disk). These are swapped (ie flushing begins) based on
memory used by the in-memory log trees and time since we last flushed
something.

The flushing facility monitors the amount of entries coming in and being
flushed out, and calibrates itself to try to flush enough each txg to
keep up with the ingest rate without competing too much with other IO.
Multiple tuneables are provided to control the flushing facility.

All the histograms and stats are update to accomodate the log as a
separate entry store. zdb gains knowledge of how to count them and dump
them. Documentation included!

Reviewed-by: Alexander Motin <mav@FreeBSD.org>
Reviewed-by: Brian Behlendorf <behlendorf1@llnl.gov>
Co-authored-by: Allan Jude <allan@klarasystems.com>
Signed-off-by: Rob Norris <rob.norris@klarasystems.com>
Sponsored-by: Klara, Inc.
Sponsored-by: iXsystems, Inc.
Closes #15895

tests/zfs-tests/include/tunables.cfg

@@ -31,6 +31,7 @@ DBUF_CACHE_SHIFT		dbuf.cache_shift		dbuf_cache_shift
 DDT_ZAP_DEFAULT_BS		dedup.ddt_zap_default_bs	ddt_zap_default_bs
 DDT_ZAP_DEFAULT_IBS		dedup.ddt_zap_default_ibs	ddt_zap_default_ibs
 DDT_DATA_IS_SPECIAL		ddt_data_is_special		zfs_ddt_data_is_special
+DEDUP_LOG_TXG_MAX		dedup.log_txg_max		zfs_dedup_log_txg_max
 DEADMAN_CHECKTIME_MS		deadman.checktime_ms		zfs_deadman_checktime_ms
 DEADMAN_EVENTS_PER_SECOND	deadman_events_per_second	zfs_deadman_events_per_second
 DEADMAN_FAILMODE		deadman.failmode		zfs_deadman_failmode

tests/zfs-tests/tests/functional/dedup/dedup_fdt_create.ksh

@@ -29,9 +29,16 @@
 
 log_assert "basic dedup (FDT) operations work"
 
+# we set the dedup log txg interval to 1, to get a log flush every txg,
+# effectively disabling the log. without this it's hard to predict when and
+# where things appear on-disk
+log_must save_tunable DEDUP_LOG_TXG_MAX
+log_must set_tunable32 DEDUP_LOG_TXG_MAX 1
+
 function cleanup
 {
 	destroy_pool $TESTPOOL
+	log_must restore_tunable DEDUP_LOG_TXG_MAX
 }
 
 log_onexit cleanup

tests/zfs-tests/tests/functional/dedup/dedup_fdt_import.ksh

@@ -29,9 +29,16 @@
 
 log_assert "dedup (FDT) retains version after import"
 
+# we set the dedup log txg interval to 1, to get a log flush every txg,
+# effectively disabling the log. without this it's hard to predict when and
+# where things appear on-disk
+log_must save_tunable DEDUP_LOG_TXG_MAX
+log_must set_tunable32 DEDUP_LOG_TXG_MAX 1
+
 function cleanup
 {
 	destroy_pool $TESTPOOL
+	log_must restore_tunable DEDUP_LOG_TXG_MAX
 }
 
 log_onexit cleanup

tests/zfs-tests/tests/functional/dedup/dedup_legacy_fdt_mixed.ksh

@@ -30,9 +30,16 @@
 
 log_assert "legacy and FDT dedup tables on the same pool can happily coexist"
 
+# we set the dedup log txg interval to 1, to get a log flush every txg,
+# effectively disabling the log. without this it's hard to predict when and
+# where things appear on-disk
+log_must save_tunable DEDUP_LOG_TXG_MAX
+log_must set_tunable32 DEDUP_LOG_TXG_MAX 1
+
 function cleanup
 {
 	destroy_pool $TESTPOOL
+	log_must restore_tunable DEDUP_LOG_TXG_MAX
 }
 
 log_onexit cleanup

tests/zfs-tests/tests/functional/dedup/dedup_legacy_fdt_upgrade.ksh

@@ -30,9 +30,16 @@
 
 log_assert "legacy dedup tables work after upgrade; new dedup tables created as FDT"
 
+# we set the dedup log txg interval to 1, to get a log flush every txg,
+# effectively disabling the log. without this it's hard to predict when and
+# where things appear on-disk
+log_must save_tunable DEDUP_LOG_TXG_MAX
+log_must set_tunable32 DEDUP_LOG_TXG_MAX 1
+
 function cleanup
 {
 	destroy_pool $TESTPOOL
+	log_must restore_tunable DEDUP_LOG_TXG_MAX
 }
 
 log_onexit cleanup

tests/zfs-tests/tests/functional/dedup/dedup_quota.ksh

@@ -51,6 +51,12 @@ POOL="dedup_pool"
 
 save_tunable TXG_TIMEOUT
 
+# we set the dedup log txg interval to 1, to get a log flush every txg,
+# effectively disabling the log. without this it's hard to predict when and
+# where things appear on-disk
+log_must save_tunable DEDUP_LOG_TXG_MAX
+log_must set_tunable32 DEDUP_LOG_TXG_MAX 1
+
 function cleanup
 {
 	if poolexists $POOL ; then
@@ -58,6 +64,7 @@ function cleanup
 	fi
 	log_must rm -fd $VDEV_GENERAL $VDEV_DEDUP $MOUNTDIR
 	log_must restore_tunable TXG_TIMEOUT
+	log_must restore_tunable DEDUP_LOG_TXG_MAX
 }
 
 
@@ -206,10 +213,15 @@ function ddt_dedup_vdev_limit
 
 	#
 	# With no DDT quota in place, the above workload will produce over
-	# 800,000 entries by using space in the normal class. With a quota,
-	# it will be well below 500,000 entries.
+	# 800,000 entries by using space in the normal class. With a quota, it
+	# should be well under 500,000. However, logged entries are hard to
+	# account for because they can appear on both logs, and can also
+	# represent an eventual removal. This isn't easily visible from
+	# outside, and even internally can result in going slightly over quota.
+	# For here, we just set the entry count a little higher than what we
+	# expect to allow for some instability.
 	#
-	log_must test $(ddt_entries) -le 500000
+	log_must test $(ddt_entries) -le 600000
 
 	do_clean
 }