This release contains performance improvements and bug fixes since the 2.22.0 release. We recommend that you upgrade at the next available opportunity.

This release blocks the ability to create new concurrent refresh policies in hierarchical continuous aggregates, as in rare cases a deadlock can occur. Concurrent refresh policies were introduced in 2.21.0 and allow users to define multiple time ranges to refresh on the same continuous aggregate, e.g. data from the last hour in policy and the last day in a second policy. Existing concurrent refresh policies on hierarchical continuous aggregates will continue to be executed. To avoid any potential deadlock, remove such existing policies and create a new policy for the full range you want to refresh, of the continuous aggregate as follows:

--- Find the job ID's of the concurrent refresh policies
SELECT * FROM timescaledb_information.jobs WHERE proc_name = 'policy_refresh_continuous_aggregate';
--- Remove the job
SELECT delete_job("<job_id_of_concurrent_policy>");
--- Create new policy for hierarchical continuous aggregate
SELECT add_continuous_aggregate_policy('<name_of_materialized_view>',
  start_offset => INTERVAL '1 month',
  end_offset => INTERVAL '1 day',
  schedule_interval => INTERVAL '1 hour');

Bugfixes

• #7766 Load OSM extension in retention background worker to drop tiered chunks
• #8550 Error in gapfill with expressions over aggregates and groupby columns and out-of-order columns
• #8593 Error on change of invalidation method for continuous aggregate
• #8599 Fix attnum mismatch bug in chunk constraint checks
• #8607 Fix interrupted continous aggregate refresh materialization phase leaving behind pending materialization ranges
• #8638 ALTER TABLE RESET for orderby settings
• #8644 Fix migration script for sparse index configuration
• #8657 Fix when using UUIDv7 partitioning

Thanks

• @MKrkkl for reporting a bug in Gapfill queries with expressions over aggregates and groupby columns
• @brandonpurcell-dev for creating a test case that showed a bug in CREATE TABLE WITH when using UUIDv7 partitioning
• @snyrkill for reporting a bug when interrupting a continous aggregate refresh