storage: integrate columnar blocks, disabled by default

This commit defines a colblk.KeySchema for use with CockroachDB keys, bumps the
format major version in 24.3 to FormatColumnarBlocks and gates use of columnar
blocks behind a new `storage.columnar_blocks.enabled` cluster setting.

This commit is port of code that's been iterated on within Pebble:
https://github.com/cockroachdb/pebble/blob/9522f39482c31a6bc956b438b92d94eb1b74ec79/internal/crdbtest/crdbtest.go

Epic: none
Release note: none

docs/generated/settings/settings-for-tenants.txt

@@ -385,6 +385,7 @@ sql.ttl.job.enabled	boolean	true	whether the TTL job is enabled	application
 sql.txn.read_committed_isolation.enabled	boolean	true	set to true to allow transactions to use the READ COMMITTED isolation level if specified by BEGIN/SET commands	application
 sql.txn.repeatable_read_isolation.enabled (alias: sql.txn.snapshot_isolation.enabled)	boolean	false	set to true to allow transactions to use the REPEATABLE READ isolation level if specified by BEGIN/SET commands	application
 sql.txn_fingerprint_id_cache.capacity	integer	100	the maximum number of txn fingerprint IDs stored	application
+storage.columnar_blocks.enabled	boolean	false	set to true to enable columnar-blocks to store KVs in a columnar format	system-visible
 storage.ingestion.value_blocks.enabled	boolean	true	set to true to enable writing of value blocks in ingestion sstables	application
 storage.max_sync_duration	duration	20s	maximum duration for disk operations; any operations that take longer than this setting trigger a warning log entry or process crash	system-visible
 storage.max_sync_duration.fatal.enabled	boolean	true	if true, fatal the process when a disk operation exceeds storage.max_sync_duration	application

docs/generated/settings/settings.html

@@ -339,6 +339,7 @@
 <tr><td><div id="setting-sql-txn-read-committed-isolation-enabled" class="anchored"><code>sql.txn.read_committed_isolation.enabled</code></div></td><td>boolean</td><td><code>true</code></td><td>set to true to allow transactions to use the READ COMMITTED isolation level if specified by BEGIN/SET commands</td><td>Serverless/Dedicated/Self-Hosted</td></tr>
 <tr><td><div id="setting-sql-txn-snapshot-isolation-enabled" class="anchored"><code>sql.txn.repeatable_read_isolation.enabled<br />(alias: sql.txn.snapshot_isolation.enabled)</code></div></td><td>boolean</td><td><code>false</code></td><td>set to true to allow transactions to use the REPEATABLE READ isolation level if specified by BEGIN/SET commands</td><td>Serverless/Dedicated/Self-Hosted</td></tr>
 <tr><td><div id="setting-sql-txn-fingerprint-id-cache-capacity" class="anchored"><code>sql.txn_fingerprint_id_cache.capacity</code></div></td><td>integer</td><td><code>100</code></td><td>the maximum number of txn fingerprint IDs stored</td><td>Serverless/Dedicated/Self-Hosted</td></tr>
+<tr><td><div id="setting-storage-columnar-blocks-enabled" class="anchored"><code>storage.columnar_blocks.enabled</code></div></td><td>boolean</td><td><code>false</code></td><td>set to true to enable columnar-blocks to store KVs in a columnar format</td><td>Dedicated/Self-hosted (read-write); Serverless (read-only)</td></tr>
 <tr><td><div id="setting-storage-experimental-eventually-file-only-snapshots-enabled" class="anchored"><code>storage.experimental.eventually_file_only_snapshots.enabled</code></div></td><td>boolean</td><td><code>true</code></td><td>set to false to disable eventually-file-only-snapshots (kv.snapshot_receiver.excise.enabled must also be false)</td><td>Dedicated/Self-Hosted</td></tr>
 <tr><td><div id="setting-storage-ingest-split-enabled" class="anchored"><code>storage.ingest_split.enabled</code></div></td><td>boolean</td><td><code>true</code></td><td>set to false to disable ingest-time splitting that lowers write-amplification</td><td>Dedicated/Self-Hosted</td></tr>
 <tr><td><div id="setting-storage-ingestion-value-blocks-enabled" class="anchored"><code>storage.ingestion.value_blocks.enabled</code></div></td><td>boolean</td><td><code>true</code></td><td>set to true to enable writing of value blocks in ingestion sstables</td><td>Serverless/Dedicated/Self-Hosted</td></tr>

go.mod

@@ -128,6 +128,7 @@ require (
 	github.com/cockroachdb/cmux v0.0.0-20170110192607-30d10be49292
 	github.com/cockroachdb/cockroach-go/v2 v2.3.7
 	github.com/cockroachdb/crlfmt v0.0.0-20221214225007-b2fc5c302548
+	github.com/cockroachdb/crlib v0.0.0-20241015224233-894974b3ad94
 	github.com/cockroachdb/datadriven v1.0.3-0.20240530155848-7682d40af056
 	github.com/cockroachdb/errors v1.11.3
 	github.com/cockroachdb/fifo v0.0.0-20240606204812-0bbfbd93a7ce
@@ -306,7 +307,6 @@ require (
 	github.com/cespare/xxhash/v2 v2.2.0 // indirect
 	github.com/charmbracelet/bubbletea v0.23.1 // indirect
 	github.com/charmbracelet/lipgloss v0.6.0 // indirect
-	github.com/cockroachdb/crlib v0.0.0-20241015224233-894974b3ad94 // indirect
 	github.com/cockroachdb/swiss v0.0.0-20240612210725-f4de07ae6964 // indirect
 	github.com/danieljoos/wincred v1.1.2 // indirect
 	github.com/decred/dcrd/dcrec/secp256k1/v4 v4.3.0 // indirect

pkg/storage/BUILD.bazel

@@ -28,6 +28,7 @@ go_library(
         "pebble.go",
         "pebble_batch.go",
         "pebble_iterator.go",
+        "pebble_key_schema.go",
         "pebble_logger_and_tracer.go",
         "pebble_merge.go",
         "pebble_mvcc_scanner.go",
@@ -86,6 +87,7 @@ go_library(
         "//pkg/util/timeutil",
         "//pkg/util/tracing",
         "//pkg/util/uuid",
+        "@com_github_cockroachdb_crlib//crbytes",
         "@com_github_cockroachdb_errors//:errors",
         "@com_github_cockroachdb_errors//oserror",
         "@com_github_cockroachdb_fifo//:fifo",
@@ -100,6 +102,7 @@ go_library(
         "@com_github_cockroachdb_pebble//replay",
         "@com_github_cockroachdb_pebble//sstable",
         "@com_github_cockroachdb_pebble//sstable/block",
+        "@com_github_cockroachdb_pebble//sstable/colblk",
         "@com_github_cockroachdb_pebble//vfs",
         "@com_github_cockroachdb_pebble//wal",
         "@com_github_cockroachdb_redact//:redact",
@@ -139,6 +142,7 @@ go_test(
         "mvcc_value_test.go",
         "open_test.go",
         "pebble_iterator_test.go",
+        "pebble_key_schema_test.go",
         "pebble_mvcc_scanner_test.go",
         "pebble_test.go",
         "read_as_of_iterator_test.go",
@@ -198,6 +202,8 @@ go_test(
         "//pkg/util/timeutil",
         "//pkg/util/uint128",
         "//pkg/util/uuid",
+        "@com_github_cockroachdb_crlib//crbytes",
+        "@com_github_cockroachdb_crlib//crstrings",
         "@com_github_cockroachdb_datadriven//:datadriven",
         "@com_github_cockroachdb_errors//:errors",
         "@com_github_cockroachdb_errors//oserror",
@@ -206,9 +212,11 @@ go_test(
         "@com_github_cockroachdb_pebble//objstorage/objstorageprovider",
         "@com_github_cockroachdb_pebble//sstable",
         "@com_github_cockroachdb_pebble//sstable/block",
+        "@com_github_cockroachdb_pebble//sstable/colblk",
         "@com_github_cockroachdb_pebble//vfs",
         "@com_github_cockroachdb_redact//:redact",
         "@com_github_kr_pretty//:pretty",
+        "@com_github_olekukonko_tablewriter//:tablewriter",
         "@com_github_stretchr_testify//assert",
         "@com_github_stretchr_testify//require",
         "@org_golang_google_protobuf//proto",

pkg/storage/pebble.go

@@ -102,6 +102,15 @@ var IngestSplitEnabled = settings.RegisterBoolSetting(
 	settings.WithPublic,
 )
 
+// columnarBlocksEnabled controls whether columnar-blocks are enabled in Pebble.
+var columnarBlocksEnabled = settings.RegisterBoolSetting(
+	settings.SystemVisible,
+	"storage.columnar_blocks.enabled",
+	"set to true to enable columnar-blocks to store KVs in a columnar format",
+	false, // TODO(jackson): Metamorphicize this.
+	settings.WithPublic,
+)
+
 // IngestAsFlushable controls whether ingested sstables that overlap the
 // memtable may be lazily ingested: written to the WAL and enqueued in the list
 // of flushables (eg, memtables, large batches and now lazily-ingested
@@ -407,6 +416,30 @@ func EngineSuffixCompare(a, b []byte) int {
 	return bytes.Compare(b[:len(b)-1], a[:len(a)-1])
 }
 
+// EnginePointSuffixCompare compares suffixes of Cockroach point keys (which are
+// composed of the version and a trailing version-length byte); the version can
+// be an MVCC timestamp or a lock key. EnginePointSuffixCompare differs from
+// EngineSuffixCompare, because EnginePointSuffixCompare normalizes the
+// suffixes. Ideally we'd have one function that implemented the semantics of
+// EnginePointSuffixCompare, but due to historical reasons, range key suffix
+// comparisons must not perform normalization.
+//
+// See https://github.com/cockroachdb/cockroach/issues/130533
+func EnginePointSuffixCompare(a, b []byte) int {
+	// NB: For performance, this routine manually splits the key into the
+	// user-key and version components rather than using DecodeEngineKey. In
+	// most situations, use DecodeEngineKey or GetKeyPartFromEngineKey or
+	// SplitMVCCKey instead of doing this.
+	if len(a) == 0 || len(b) == 0 {
+		// Empty suffixes sort before non-empty suffixes.
+		return cmp.Compare(len(a), len(b))
+	}
+	return bytes.Compare(
+		normalizeEngineSuffixForCompare(b),
+		normalizeEngineSuffixForCompare(a),
+	)
+}
+
 func checkEngineKey(k []byte) {
 	if len(k) == 0 {
 		panic(errors.AssertionFailedf("empty key"))
@@ -792,8 +825,9 @@ const MinimumSupportedFormatVersion = pebble.FormatSyntheticPrefixSuffix
 // DefaultPebbleOptions returns the default pebble options.
 func DefaultPebbleOptions() *pebble.Options {
 	opts := &pebble.Options{
-		Comparer: EngineComparer,
-		FS:       vfs.Default,
+		Comparer:  EngineComparer,
+		FS:        vfs.Default,
+		KeySchema: keySchema,
 		// A value of 2 triggers a compaction when there is 1 sub-level.
 		L0CompactionThreshold: 2,
 		L0StopWritesThreshold: 1000,
@@ -1190,6 +1224,9 @@ func newPebble(ctx context.Context, cfg engineConfig) (p *Pebble, err error) {
 	cfg.opts.Experimental.IngestSplit = func() bool {
 		return IngestSplitEnabled.Get(&cfg.settings.SV)
 	}
+	cfg.opts.Experimental.EnableColumnarBlocks = func() bool {
+		return columnarBlocksEnabled.Get(&cfg.settings.SV)
+	}
 
 	auxDir := cfg.opts.FS.PathJoin(cfg.env.Dir, base.AuxiliaryDir)
 	if !cfg.env.IsReadOnly() {
@@ -2512,7 +2549,7 @@ func (p *Pebble) CreateCheckpoint(dir string, spans []roachpb.Span) error {
 // version associated with it, since they did so during the fence version.
 var pebbleFormatVersionMap = map[clusterversion.Key]pebble.FormatMajorVersion{
 	clusterversion.V24_1: pebble.FormatSyntheticPrefixSuffix,
-	clusterversion.V24_3: pebble.FormatFlushableIngestExcises,
+	clusterversion.V24_3: pebble.FormatColumnarBlocks,
 }
 
 // pebbleFormatVersionKeys contains the keys in the map above, in descending order.