Export from a node

Exporting (or querying) data can be done by running a pipeline that begins with the export source. When managing a pipeline through the app or the API, all pipeline operators run within the node. When using the CLI, at least the export operator runs within the node.

Let's bring back a sample of historical data we imported in the previous section:

export
head

Think of export being the entire data at a node. As this can grow quickly, you may query only subsets of it, e.g., by filtering using where:

export
where orig_bytes < 1 KiB

Logically, this query would first export the entire historical data, and then begin filtering the data. But since Tenzir does predicate pushdown, the pipeline executor will analyze the query and push the expression in where with the predicate orig_bytes < 1 KiB "down" to the export operator. Tenzir's storage engine then asks its catalog to identify the relevant subset of partitions that the query should execute on. This dramatically improves the query performance for selective workloads, such as point queries for single values or specific time ranges.

To figure out the shape of the data to query, you can show available schemas.