Catalog Metadata

Every table makes metadata about itself available through special virtual tables that you can query through Hydrolix's SQL API or query UI. These include catalogs of the files and directories that hold tables' data and indexes, and descriptions of the columns and time-intervals held by each table's constituent partitions.

Catalog

A catalog virtual table contains one row for every partition used by its associated fluid table. The row's data include that partition's filesystem path, the span of time-series data it contains, and the amount of storage space it takes up.

Catalog columns

In the following table, "this partition" means the particular partition of fluid table data that each catalog row describes.

NamePurpose
partitionThis partition's filesystem path, relative to its table's cloud-storage location.
min_timestampThe earliest timestamp of the contiguous time-series data held by this partition.
max_timestampThe latest timestamp of the contiguous time-series data held by this partition.
manifest_sizeThe size of this partition's manifest file, in bytes.
data_sizeThe size of this partition's data file, in bytes.
index_sizeThe size of this partition's index file, in bytes.

Querying a table's catalog

To query a table's catalog, refer to it in a query as TABLE-NAME#.catalog.

For example, to view the entire catalog of the table my-data in the project my-project:

SELECT * FROM `my-project`.`my-data#.catalog`

To view the total size, in bytes, of the table my-data:

SELECT sum(manifest_size + data_size + index_size) as total_size
    FROM `my-project`.`my-data#.catalog`

👍

Useful for Partition Sizing

Catalog access is really useful to determine the size of partitions being stored within your cluster. Each row of data is automatically stored within a partition, to ensure performance Hydrolix recommends aiming for partition sizes of between 1 and 1.5 GB.

A good query for this is shown below.

select toStartOfDay(min_timestamp) as day,
    count(min_timestamp) as count,
    min(min_timestamp) as min_time,
    max(max_timestamp) as max_time,
    formatReadableSize(avg(manifest_size)) as manifestSize,
    formatReadableSize(avg(data_size)) as dataSize,
    formatReadableSize(avg(index_size)) as indexSize,
    formatReadableSize(min(manifest_size + data_size + index_size)) as minTotal,
    formatReadableSize(median(manifest_size + data_size + index_size)) as medianTotal,
    formatReadableSize(max(manifest_size + data_size + index_size)) as maxTotal,
    formatReadableSize(avg(manifest_size + data_size + index_size)) as avgTotal,
    formatReadableSize(sum(manifest_size + data_size + index_size)) as sumTotal
from project.`table#.catalog`
group by day
order by day

Metadata

A metadata virtual table contains several rows for every partition used by its associated fluid table, with one row for every column handled by that partition. In other words, every row in this virtual table represents one partition-and-column tuple from the files that hold your table's data.

Because a Hydrolix table can have multiple transform schemas associated with it, the partitions of a given table might contain different lists of columns from one another. Querying a table's metadata lets you see exactly which columns of data each of its partitions handle, and how they handle them.

Metadata columns

In the following table, "this partition" means the particular partition of fluid table data that each metadata row describes, and "this column" means the specific column within that partition that the metadata row describes.

NamePurpose
partitionThis partition's filesystem path, relative to its table's cloud-storage location.
min_timestampThe earliest timestamp of the contiguous time-series data held by this partition.
max_timestampThe latest timestamp of the contiguous time-series data held by this partition.
column_nameThe name of this column.
column_index1 if Hydrolix treats this column's data as indexible; 0 otherwise.
column_type
column_unique_values
column_order
dict_compressed_bytes
dict_uncompressed_bytes

Querying a table's metadata

To query a table's metadata, refer to it in a query as TABLE-NAME#.metadata.

For example, to view all the metadata of the table my-data in the project my-project:

SELECT * FROM `my-project`.`my-data#.metadata`

And this query would list all the columns that a given table has ever handled, across all its partitions, including the earliest and latest times each was used:

SELECT column_name, min(min_timestamp) AS first_appearance, max(max_timestamp) AS last_appearance
    FROM `my-project`.`my-data#.metadata`
    GROUP BY column_name
    ORDER BY first_appearance, column_name