*: audit all duration settings to add validation function

This commit adjusts the DurationSetting so that if no validation
function is provided, it assumes the non-negative duration.

This commit audits all duration settings to add validation functions. In
vast majority cases non-negative duration seems reasonable, for three
I'm adding positive validation:
- `kv.gc.txn_cleanup_threshold`
- `jobs.scheduler.pace`
- `sql.stats.max_timestamp_age`.

`sql.crdb_internal.table_row_statistics.as_of_time` got a non-positive
validation.

Note that since we're adding new validation logic, it can be possible
for existing setting values to not pass the validation anymore. In such
case the default will be used instead (and a log message will be
written) which seems like a reasonable behavior.

Release note: None

Author: yuzefovich

pkg/backup/backup_processor.go

@@ -53,21 +53,18 @@ var (
 		"bulkio.backup.read_with_priority_after",
 		"amount of time since the read-as-of time above which a BACKUP should use priority when retrying reads",
 		time.Minute,
-		settings.NonNegativeDuration,
 		settings.WithPublic)
 	delayPerAttempt = settings.RegisterDurationSetting(
 		settings.ApplicationLevel,
 		"bulkio.backup.read_retry_delay",
 		"amount of time since the read-as-of time, per-prior attempt, to wait before making another attempt",
 		time.Second*5,
-		settings.NonNegativeDuration,
 	)
 	timeoutPerAttempt = settings.RegisterDurationSetting(
 		settings.ApplicationLevel,
 		"bulkio.backup.read_timeout",
 		"amount of time after which a read attempt is considered timed out, which causes the backup to fail",
 		time.Minute*5,
-		settings.NonNegativeDuration,
 		settings.WithPublic)
 
 	preSplitExports = settings.RegisterBoolSetting(

pkg/ccl/changefeedccl/cdcevent/rowfetcher_cache.go

@@ -38,7 +38,6 @@ var traceKVLogFrequency = settings.RegisterDurationSetting(
 	"changefeed.cdcevent.trace_kv.log_frequency",
 	"controls how frequently KVs are logged when KV tracing is enabled",
 	500*time.Millisecond,
-	settings.NonNegativeDuration,
 )
 
 // rowFetcherCache maintains a cache of single table row.Fetchers. Given a key

pkg/ccl/changefeedccl/changefeed_processors.go

@@ -694,7 +694,6 @@ var aggregatorHeartbeatFrequency = settings.RegisterDurationSetting(
 	"changefeed aggregator will emit a heartbeat message to the coordinator with this frequency; 0 disables. "+
 		"The setting value should be <=1/2 of server.shutdown.jobs.timeout period",
 	4*time.Second,
-	settings.NonNegativeDuration,
 )
 
 var aggregatorFlushJitter = settings.RegisterFloatSetting(

pkg/ccl/changefeedccl/changefeedbase/settings.go

@@ -23,7 +23,6 @@ var TableDescriptorPollInterval = settings.RegisterDurationSetting(
 	"changefeed.experimental_poll_interval",
 	"polling interval for the table descriptors",
 	1*time.Second,
-	settings.NonNegativeDuration,
 )
 
 // DefaultMinCheckpointFrequency is the default frequency to flush sink.
@@ -53,7 +52,6 @@ var SlowSpanLogThreshold = settings.RegisterDurationSetting(
 	"changefeed.slow_span_log_threshold",
 	"a changefeed will log spans with resolved timestamps this far behind the current wall-clock time; if 0, a default value is calculated based on other cluster settings",
 	0,
-	settings.NonNegativeDuration,
 )
 
 // IdleTimeout controls how long the changefeed will wait for a new KV being
@@ -63,7 +61,6 @@ var IdleTimeout = settings.RegisterDurationSetting(
 	"changefeed.idle_timeout",
 	"a changefeed will mark itself idle if no changes have been emitted for greater than this duration; if 0, the changefeed will never be marked idle",
 	10*time.Minute,
-	settings.NonNegativeDuration,
 	settings.WithName("changefeed.auto_idle.timeout"),
 )
 
@@ -75,7 +72,6 @@ var SpanCheckpointInterval = settings.RegisterDurationSetting(
 	"interval at which span-level checkpoints will be written; "+
 		"if 0, span-level checkpoints are disabled",
 	10*time.Minute,
-	settings.NonNegativeDuration,
 	settings.WithName("changefeed.span_checkpoint.interval"),
 )
 
@@ -90,7 +86,6 @@ var SpanCheckpointLagThreshold = settings.RegisterDurationSetting(
 		"to save leading span progress is written; if 0, span-level checkpoints "+
 		"due to lagging spans is disabled",
 	10*time.Minute,
-	settings.NonNegativeDuration,
 	settings.WithPublic,
 	settings.WithName("changefeed.span_checkpoint.lag_threshold"),
 )
@@ -185,7 +180,6 @@ var ResolvedTimestampMinUpdateInterval = settings.RegisterDurationSetting(
 		"updated again; default of 0 means no minimum interval is enforced but "+
 		"updating will still be limited by the average time it takes to checkpoint progress",
 	0,
-	settings.NonNegativeDuration,
 	settings.WithPublic,
 	settings.WithName("changefeed.resolved_timestamp.min_update_interval"),
 )
@@ -227,7 +221,6 @@ var MaxProtectedTimestampAge = settings.RegisterDurationSetting(
 	"changefeed.protect_timestamp.max_age",
 	"fail the changefeed if the protected timestamp age exceeds this threshold; 0 disables expiration",
 	4*24*time.Hour,
-	settings.NonNegativeDuration,
 	settings.WithPublic)
 
 // BatchReductionRetryEnabled enables the temporary reduction of batch sizes upon kafka message too large errors
@@ -354,5 +347,4 @@ var Quantize = settings.RegisterDurationSettingWithExplicitUnit(
 	"changefeed.resolved_timestamp.granularity",
 	"the granularity at which changefeed progress are quantized to make tracking more efficient",
 	0,
-	settings.NonNegativeDuration,
 )

pkg/ccl/changefeedccl/telemetry.go

@@ -181,5 +181,4 @@ var continuousTelemetryInterval = settings.RegisterDurationSetting(
 	"determines the interval at which each node emits continuous telemetry events"+
 		" during the lifespan of every changefeed; setting a zero value disables logging",
 	24*time.Hour,
-	settings.NonNegativeDuration,
 )

pkg/ccl/jwtauthccl/settings.go

@@ -142,7 +142,6 @@ var JWTAuthClientTimeout = settings.RegisterDurationSetting(
 	"sets the client timeout for external calls made during JWT authentication "+
 		"(e.g. fetching JWKS, etc.)",
 	15*time.Second,
-	settings.NonNegativeDuration,
 	settings.WithPublic,
 )
 

pkg/ccl/kvccl/kvfollowerreadsccl/followerreads.go

@@ -39,7 +39,6 @@ var ClosedTimestampPropagationSlack = settings.RegisterDurationSetting(
 		"propagate from a leaseholder to followers. This is taken into account by "+
 		"follower_read_timestamp().",
 	time.Second,
-	settings.NonNegativeDuration,
 )
 
 // getFollowerReadLag returns the (negative) offset duration from hlc.Now()

pkg/ccl/oidcccl/settings.go

@@ -81,7 +81,6 @@ var OIDCAuthClientTimeout = settings.RegisterDurationSetting(
 	"sets the client timeout for external calls made during OIDC authentication "+
 		"(e.g. authorization code flow, etc.)",
 	15*time.Second,
-	settings.NonNegativeDuration,
 	settings.WithPublic,
 )
 

pkg/crosscluster/logical/logical_replication_job.go

@@ -57,8 +57,7 @@ var (
 		settings.ApplicationLevel,
 		"logical_replication.consumer.job_checkpoint_frequency",
 		"controls the frequency with which the job updates their progress; if 0, disabled",
-		10*time.Second,
-		settings.NonNegativeDuration)
+		10*time.Second)
 
 	// heartbeatFrequency controls frequency the stream replication
 	// destination cluster sends heartbeat to the source cluster to keep
@@ -69,7 +68,6 @@ var (
 		"controls frequency the stream replication destination cluster sends heartbeat "+
 			"to the source cluster to keep the stream alive",
 		30*time.Second,
-		settings.NonNegativeDuration,
 	)
 )
 

pkg/crosscluster/physical/stream_ingestion_processor.go

@@ -88,7 +88,6 @@ var cutoverSignalPollInterval = settings.RegisterDurationSetting(
 	"bulkio.stream_ingestion.failover_signal_poll_interval",
 	"the interval at which the stream ingestion job checks if it has been signaled to cutover",
 	10*time.Second,
-	settings.NonNegativeDuration,
 	settings.WithName("physical_replication.consumer.failover_signal_poll_interval"),
 )