sql: introduce contention event store

This commit introduces contention event store, an in-memory key-value
FIFO store that stores historical contention events and map collected
contention events to their corresponding transaction fingerprint IDs.

Release note (sql change): A new cluster setting
`sql.contention.event_store.capacity` is added . This cluster setting
can be used to control the in-memory capacity of contention event store.
When this setting is set to zero, the contention event store is
disabled.

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

@@ -78,6 +78,7 @@ server.web_session.purge.max_deletions_per_cycle	integer	10	the maximum number o
 server.web_session.purge.period	duration	1h0m0s	the time until old sessions are deleted
 server.web_session.purge.ttl	duration	1h0m0s	if nonzero, entries in system.web_sessions older than this duration are periodically purged
 server.web_session_timeout	duration	168h0m0s	the duration that a newly created web session will be valid
+sql.contention.event_store.capacity	byte size	64 MiB	the in-memory storage capacity per-node of contention event store
 sql.contention.txn_id_cache.max_size	byte size	64 MiB	the maximum byte size TxnID cache will use (set to 0 to disable)
 sql.cross_db_fks.enabled	boolean	false	if true, creating foreign key references across databases is allowed
 sql.cross_db_sequence_owners.enabled	boolean	false	if true, creating sequences owned by tables from other databases is allowed

docs/generated/settings/settings.html

@@ -90,6 +90,7 @@
 <tr><td><code>server.web_session.purge.period</code></td><td>duration</td><td><code>1h0m0s</code></td><td>the time until old sessions are deleted</td></tr>
 <tr><td><code>server.web_session.purge.ttl</code></td><td>duration</td><td><code>1h0m0s</code></td><td>if nonzero, entries in system.web_sessions older than this duration are periodically purged</td></tr>
 <tr><td><code>server.web_session_timeout</code></td><td>duration</td><td><code>168h0m0s</code></td><td>the duration that a newly created web session will be valid</td></tr>
+<tr><td><code>sql.contention.event_store.capacity</code></td><td>byte size</td><td><code>64 MiB</code></td><td>the in-memory storage capacity per-node of contention event store</td></tr>
 <tr><td><code>sql.contention.txn_id_cache.max_size</code></td><td>byte size</td><td><code>64 MiB</code></td><td>the maximum byte size TxnID cache will use (set to 0 to disable)</td></tr>
 <tr><td><code>sql.cross_db_fks.enabled</code></td><td>boolean</td><td><code>false</code></td><td>if true, creating foreign key references across databases is allowed</td></tr>
 <tr><td><code>sql.cross_db_sequence_owners.enabled</code></td><td>boolean</td><td><code>false</code></td><td>if true, creating sequences owned by tables from other databases is allowed</td></tr>

pkg/server/server_sql.go

@@ -660,6 +660,8 @@ func newSQLServer(ctx context.Context, cfg sqlServerArgs) (*SQLServer, error) {
 	if hasNodeLiveness {
 		traceCollector = collector.New(cfg.nodeDialer, nodeLiveness, cfg.Tracer)
 	}
+	contentionRegistry := contention.NewRegistry(cfg.Settings, cfg.sqlStatusServer.TxnIDResolution)
+	contentionRegistry.Start(ctx, cfg.stopper)
 
 	*execCfg = sql.ExecutorConfig{
 		Settings:                cfg.Settings,
@@ -682,7 +684,7 @@ func newSQLServer(ctx context.Context, cfg sqlServerArgs) (*SQLServer, error) {
 		SQLStatusServer:         cfg.sqlStatusServer,
 		RegionsServer:           cfg.regionsServer,
 		SessionRegistry:         cfg.sessionRegistry,
-		ContentionRegistry:      contention.NewRegistry(),
+		ContentionRegistry:      contentionRegistry,
 		SQLLiveness:             cfg.sqlLivenessProvider,
 		JobRegistry:             jobRegistry,
 		VirtualSchemas:          virtualSchemas,

pkg/sql/contention/BUILD.bazel

@@ -3,6 +3,8 @@ load("@io_bazel_rules_go//go:def.bzl", "go_library", "go_test")
 go_library(
     name = "contention",
     srcs = [
+        "cluster_settings.go",
+        "event_store.go",
         "registry.go",
         "resolver.go",
         "test_utils.go",
@@ -13,10 +15,16 @@ go_library(
         "//pkg/keys",
         "//pkg/roachpb",
         "//pkg/server/serverpb",
+        "//pkg/settings",
+        "//pkg/settings/cluster",
         "//pkg/sql/catalog/descpb",
+        "//pkg/sql/contention/contentionutils",
         "//pkg/sql/contentionpb",
         "//pkg/util/cache",
+        "//pkg/util/log",
+        "//pkg/util/stop",
         "//pkg/util/syncutil",
+        "//pkg/util/timeutil",
         "//pkg/util/uuid",
         "@com_github_biogo_store//llrb",
         "@com_github_cockroachdb_errors//:errors",
@@ -27,6 +35,7 @@ go_test(
     name = "contention_test",
     size = "small",
     srcs = [
+        "event_store_test.go",
         "registry_test.go",
         "resolver_test.go",
         "utils_test.go",
@@ -36,13 +45,15 @@ go_test(
     deps = [
         "//pkg/keys",
         "//pkg/roachpb",
+        "//pkg/settings/cluster",
         "//pkg/sql/contentionpb",
         "//pkg/storage/enginepb",
         "//pkg/testutils",
         "//pkg/util/cache",
         "//pkg/util/encoding",
         "//pkg/util/leaktest",
         "//pkg/util/randutil",
+        "//pkg/util/stop",
         "//pkg/util/uuid",
         "@com_github_cockroachdb_datadriven//:datadriven",
         "@com_github_cockroachdb_errors//:errors",

pkg/sql/contention/cluster_settings.go

@@ -0,0 +1,35 @@
+// Copyright 2022 The Cockroach Authors.
+//
+// Use of this software is governed by the Business Source License
+// included in the file licenses/BSL.txt.
+//
+// As of the Change Date specified in that file, in accordance with
+// the Business Source License, use of this software will be governed
+// by the Apache License, Version 2.0, included in the file
+// licenses/APL.txt.
+
+package contention
+
+import (
+	"time"
+
+	"github.com/cockroachdb/cockroach/pkg/settings"
+)
+
+// TxnIDResolutionInterval is the cluster setting that controls how often the
+// Transaction ID Resolution is performed.
+var TxnIDResolutionInterval = settings.RegisterDurationSetting(
+	settings.TenantWritable,
+	"sql.contention.event_store.resolution_interval",
+	"the interval at which transaction ID resolution is performed (set to 0 to disable)",
+	time.Second*30,
+)
+
+// StoreCapacity is the cluster setting that controls the
+// maximum size of the contention event store.
+var StoreCapacity = settings.RegisterByteSizeSetting(
+	settings.TenantWritable,
+	"sql.contention.event_store.capacity",
+	"the in-memory storage capacity per-node of contention event store",
+	64*1024*1024, // 64 MB per node.
+).WithPublic()