ContextQMD
Back to Materialize

Materialize v0.37

doc/user/content/releases/v0.37.md

1233.0 KB
Original Source

v0.37.0

  • Add support for connecting to Confluent Schema Registry using an SSH tunnel connection to an SSH bastion server.

  • Add mz_internal.mz_cluster_replica_utilization to the system catalog. This view allows you to monitor the resource utilization for all extant cluster replicas as a % of the total resource allocation:

    mzsql
    SELECT * FROM mz_internal.mz_cluster_replica_utilization;
    nofmt
      replica_id | process_id |     cpu_percent      |    memory_percent
 ------------+------------+----------------------+----------------------
  1          | 0          |            9.6629961 |   0.6772994995117188
  2          | 0          |  0.10735560000000001 |   0.3876686096191406
  3          | 0          |           46.8730398 |      0.7110595703125

    It's important to note that these tables are part of an unstable interface of Materialize (mz_internal), which means that their values may change at any time, and you should not rely on them for tasks like capacity planning for the time being.

  • Add mz_internal.mz_storage_host_metrics and mz_internal.mz_storage_host_sizes to the system catalog. These objects respectively expose the last known CPU and RAM utilization statistics for all processes of all extant storage hosts, and a mapping of logical sizes (e.g. xlarge) to the number of processes, as well as CPU and memory allocations for each process.

    The concept of a storage host is not user-facing, and is intentionally undocumented. It refers to the physical resource allocation on which Materialize can schedule multiple sources and sinks behind the scenes.

  • Rename system catalog objects to adhere to the naming conventions defined in the Materialize SQL style guide. The affected objects are:

    SchemaOld nameNew name
    mz_internalmz_source_statusmz_source_statuses
    mz_internalmz_sink_statusmz_sink_statuses

  • Support JSON as an output format of EXPLAIN TIMESTAMP.

  • Fix a bug where the to_timestamp function would truncate the fractional part of negative timestamps (i.e. prior to the Unix epoch: January 1st, 1970 at 00:00:00 UTC) {{% gh 16610 %}}, and return an error instead of NULL when the timestamp is out of range. The new behavior matches PostgreSQL.

  • Private preview. Add a WebSocket API endpoint which supports interactive SQL queries over WebSockets.

  • Change the JSON serialization for rows emitted by the HTTP API endpoint to exactly match the JSON serialization used by FORMAT JSON Kafka sinks. Previously, the HTTP SQL endpoint serialized datums using slightly different rules.