Module insert 

Modules

otlp 🔒

Side-effect-free destination validation for prepared OTLP batches. Catalog failures propagate; row errors let the converter reject complete source-point fan-outs.

Structs

CreateAlterTableResult 🔒

Result of create_or_alter_tables_on_demand.

FlowMirrorTask 🔒

Inserter

InstantAndNormalInsertRequests

Split insert requests into normal and instant requests.

MetricInsertPreflight

MirrorLog 🔒

PreparedMetricInserts

Metric writes bound to the same resolved tables used by their exemplars.

Enums

AutoCreateTableType

Hint for the table type to create automatically.

Constants

MAX_MIRROR_PENDING_ROWS 🔒

Upper bound of rows buffered by detached flow mirror tasks on this frontend.

MIRROR_LOG_INTERVAL 🔒

How often at most each mirror event is reported.

MIRROR_LOG_NEVER_REPORTED 🔒

SPLUNK_PK_METADATA_ORDER_KEY

QueryContext extension key the Splunk HEC handler sets (to "true") on its identity path to request metadata-first primary-key ordering at table creation. It is absent for user-supplied pipelines, so their primary-key order is left untouched.

Statics

MIRROR_DROP_LOG 🔒

MIRROR_FAILURE_LOG 🔒

Functions

admit_row_insert_batches

Admits all database totals before dispatching any batch of a finite request. Each batch keeps its own protocol options and target database.

admit_write

Admits a finite request before it is split into internal writes. The returned context preserves accounting while preventing a second row debit.

apply_per_table_semantic_options

Folds the semantic keys of the table being created into table_options.

build_create_table_expr

convert_rows_time_unit 🔒

convert_timestamp_value_data 🔒

count_insert_rows 🔒

fill_table_options_for_create

Fill table options for a new table by create type.

mirror_drop_log_millis 🔒

Milliseconds since the first call, so the drop log is rate limited on a monotonic clock instead of the wall clock, which can jump.

parse_per_table_semantic_index

Parses the per-table semantic index off the context extension. Call once per create-planning round: a first write creating N tables would otherwise re-parse the whole index N times. None when the request carries no index (logs, traces, Prom RW v1) or it fails to parse.

region_inserts_rows 🔒

Rows of row data carried by inserts. A request without a rows payload carries nothing to keep buffered and contributes zero.

release_mirror_pending_rows 🔒

Releases rows of the pending mirror budget.

reorder_splunk_primary_keys 🔒

Moves Splunk’s metadata tags (host, source, sourcetype) to the front of the primary key, keeping the relative order of the remaining tags.

request_is_native_histogram 🔒

reserve_mirror_pending_rows 🔒

Reserves rows of the pending mirror budget and returns the resulting pending row count. The caller must release_mirror_pending_rows exactly these rows once they are no longer buffered, whether it spawns the batch or drops it.

table_is_native_histogram 🔒

table_time_index_unit 🔒

Returns the table’s time index unit, if any. A metric table without a timestamp column is left alone by the unit alignment: the metric engine rejects it anyway.

validate_column_count_match 🔒

validate_exemplar_table

Validates the fixed exemplar contract before writing or scanning a shared table.

validate_trace_table_model

Rejects writes from a different built-in trace model before schema mutation. Unstamped, explicitly created tables remain subject to normal schema validation.

Type Aliases

InserterRef

PerTableSemanticIndex

The parsed per-table semantic index: {schema -> {table -> {key -> value}}}, produced by the OTLP metrics encode path (where one metric can fan out into several tables with distinct keys) and the Prometheus remote write v2 path (where per-series metadata declares type/unit, and a series may override its target schema).