Merge #76726

76726: ttljob: paginate ranges when doing TTL r=rafiss a=otan

This commit batches ranges instead of scanning them in all at once. The
range batch size is controlled by the cluster setting
`sql.ttl.range_batch_size`. I don't anticipate this needing to be
controlled per table.

Release note: None

Co-authored-by: Oliver Tan <[email protected]>

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

@@ -173,6 +173,7 @@ sql.ttl.default_delete_batch_size	integer	100	default amount of rows to delete i
 sql.ttl.default_delete_rate_limit	integer	0	default delete rate limit for all TTL jobs. Use 0 to signify no rate limit.
 sql.ttl.default_range_concurrency	integer	1	default amount of ranges to process at once during a TTL delete
 sql.ttl.default_select_batch_size	integer	500	default amount of rows to select in a single query during a TTL job
+sql.ttl.range_batch_size	integer	100	amount of ranges to fetch at a time for a table during the TTL job
 timeseries.storage.enabled	boolean	true	if set, periodic timeseries data is stored within the cluster; disabling is not recommended unless you are storing the data elsewhere
 timeseries.storage.resolution_10s.ttl	duration	240h0m0s	the maximum age of time series data stored at the 10 second resolution. Data older than this is subject to rollup and deletion.
 timeseries.storage.resolution_30m.ttl	duration	2160h0m0s	the maximum age of time series data stored at the 30 minute resolution. Data older than this is subject to deletion.

docs/generated/settings/settings.html

@@ -186,6 +186,7 @@
 <tr><td><code>sql.ttl.default_delete_rate_limit</code></td><td>integer</td><td><code>0</code></td><td>default delete rate limit for all TTL jobs. Use 0 to signify no rate limit.</td></tr>
 <tr><td><code>sql.ttl.default_range_concurrency</code></td><td>integer</td><td><code>1</code></td><td>default amount of ranges to process at once during a TTL delete</td></tr>
 <tr><td><code>sql.ttl.default_select_batch_size</code></td><td>integer</td><td><code>500</code></td><td>default amount of rows to select in a single query during a TTL job</td></tr>
+<tr><td><code>sql.ttl.range_batch_size</code></td><td>integer</td><td><code>100</code></td><td>amount of ranges to fetch at a time for a table during the TTL job</td></tr>
 <tr><td><code>timeseries.storage.enabled</code></td><td>boolean</td><td><code>true</code></td><td>if set, periodic timeseries data is stored within the cluster; disabling is not recommended unless you are storing the data elsewhere</td></tr>
 <tr><td><code>timeseries.storage.resolution_10s.ttl</code></td><td>duration</td><td><code>240h0m0s</code></td><td>the maximum age of time series data stored at the 10 second resolution. Data older than this is subject to rollup and deletion.</td></tr>
 <tr><td><code>timeseries.storage.resolution_30m.ttl</code></td><td>duration</td><td><code>2160h0m0s</code></td><td>the maximum age of time series data stored at the 30 minute resolution. Data older than this is subject to deletion.</td></tr>

pkg/sql/ttl/ttljob/BUILD.bazel

@@ -13,7 +13,6 @@ go_library(
         "//pkg/jobs/jobspb",
         "//pkg/keys",
         "//pkg/kv",
-        "//pkg/kv/kvclient",
         "//pkg/roachpb",
         "//pkg/settings",
         "//pkg/settings/cluster",

pkg/sql/ttl/ttljob/ttljob.go

@@ -20,7 +20,6 @@ import (
 	"github.com/cockroachdb/cockroach/pkg/jobs/jobspb"
 	"github.com/cockroachdb/cockroach/pkg/keys"
 	"github.com/cockroachdb/cockroach/pkg/kv"
-	"github.com/cockroachdb/cockroach/pkg/kv/kvclient"
 	"github.com/cockroachdb/cockroach/pkg/roachpb"
 	"github.com/cockroachdb/cockroach/pkg/settings"
 	"github.com/cockroachdb/cockroach/pkg/settings/cluster"
@@ -70,6 +69,14 @@ var (
 		0,
 		settings.NonNegativeInt,
 	).WithPublic()
+
+	rangeBatchSize = settings.RegisterIntSetting(
+		settings.TenantWritable,
+		"sql.ttl.range_batch_size",
+		"amount of ranges to fetch at a time for a table during the TTL job",
+		100,
+		settings.PositiveInt,
+	).WithPublic()
 )
 
 type rowLevelTTLResumer struct {
@@ -228,8 +235,8 @@ func (t rowLevelTTLResumer) Resume(ctx context.Context, execCtx interface{}) err
 	var pkColumns []string
 	var pkTypes []*types.T
 	var pkDirs []descpb.IndexDescriptor_Direction
-	var ranges []kv.KeyValue
 	var name string
+	var rangeSpan roachpb.Span
 	if err := db.Txn(ctx, func(ctx context.Context, txn *kv.Txn) error {
 		desc, err := descs.GetImmutableTableByID(
 			ctx,
@@ -263,12 +270,6 @@ func (t rowLevelTTLResumer) Resume(ctx context.Context, execCtx interface{}) err
 		if ttl == nil {
 			return errors.Newf("unable to find TTL on table %s", desc.GetName())
 		}
-		ttlSettings = *ttl
-
-		ranges, err = kvclient.ScanMetaKVs(ctx, txn, desc.TableSpan(p.ExecCfg().Codec))
-		if err != nil {
-			return err
-		}
 
 		_, dbDesc, err := descs.GetImmutableDatabaseByID(
 			ctx,
@@ -299,6 +300,8 @@ func (t rowLevelTTLResumer) Resume(ctx context.Context, execCtx interface{}) err
 			tree.Name(desc.GetName()),
 		)
 		name = tn.FQString()
+		rangeSpan = desc.TableSpan(p.ExecCfg().Codec)
+		ttlSettings = *ttl
 		return nil
 	}); err != nil {
 		return err
@@ -362,20 +365,56 @@ func (t rowLevelTTLResumer) Resume(ctx context.Context, execCtx interface{}) err
 			close(ch)
 			retErr = errors.CombineErrors(retErr, g.Wait())
 		}()
-		for _, r := range ranges {
-			if err := r.ValueProto(&rangeDesc); err != nil {
-				return err
-			}
-			var nextRange rangeToProcess
-			nextRange.startPK, err = keyToDatums(rangeDesc.StartKey, p.ExecCfg().Codec, pkTypes, pkDirs, &alloc)
-			if err != nil {
+		done := false
+
+		batchSize := rangeBatchSize.Get(p.ExecCfg().SV())
+		for !done {
+			var ranges []kv.KeyValue
+
+			// Scan ranges up to rangeBatchSize.
+			if err := db.Txn(ctx, func(ctx context.Context, txn *kv.Txn) error {
+				metaStart := keys.RangeMetaKey(keys.MustAddr(rangeSpan.Key).Next())
+				metaEnd := keys.RangeMetaKey(keys.MustAddr(rangeSpan.EndKey))
+
+				kvs, err := txn.Scan(ctx, metaStart, metaEnd, batchSize)
+				if err != nil {
+					return err
+				}
+				if len(kvs) < int(batchSize) {
+					done = true
+					if len(kvs) == 0 || !kvs[len(kvs)-1].Key.Equal(metaEnd.AsRawKey()) {
+						// Normally we need to scan one more KV because the ranges are addressed by
+						// the end key.
+						extraKV, err := txn.Scan(ctx, metaEnd, keys.Meta2Prefix.PrefixEnd(), 1 /* one result */)
+						if err != nil {
+							return err
+						}
+						kvs = append(kvs, extraKV[0])
+					}
+				}
+				ranges = kvs
+				return nil
+			}); err != nil {
 				return err
 			}
-			nextRange.endPK, err = keyToDatums(rangeDesc.EndKey, p.ExecCfg().Codec, pkTypes, pkDirs, &alloc)
-			if err != nil {
-				return err
+
+			// Send these to each goroutine worker.
+			for _, r := range ranges {
+				if err := r.ValueProto(&rangeDesc); err != nil {
+					return err
+				}
+				rangeSpan.Key = rangeDesc.EndKey.AsRawKey()
+				var nextRange rangeToProcess
+				nextRange.startPK, err = keyToDatums(rangeDesc.StartKey, p.ExecCfg().Codec, pkTypes, pkDirs, &alloc)
+				if err != nil {
+					return err
+				}
+				nextRange.endPK, err = keyToDatums(rangeDesc.EndKey, p.ExecCfg().Codec, pkTypes, pkDirs, &alloc)
+				if err != nil {
+					return err
+				}
+				ch <- nextRange
 			}
-			ch <- nextRange
 		}
 		return nil
 	}(); err != nil {

pkg/sql/ttl/ttljob/ttljob_test.go

@@ -329,7 +329,11 @@ func TestRowLevelTTLJobRandomEntries(t *testing.T) {
 			})
 			defer cleanupFunc()
 
+			rangeBatchSize := 1 + rng.Intn(3)
+			t.Logf("range batch size: %d", rangeBatchSize)
+
 			th.sqlDB.Exec(t, tc.createTable)
+			th.sqlDB.Exec(t, `SET CLUSTER SETTING sql.ttl.range_batch_size = $1`, rangeBatchSize)
 
 			// Extract the columns from CREATE TABLE.
 			stmt, err := parser.ParseOne(tc.createTable)