Add compact_chunk function - #9957
Merged
Merged
Conversation
antekresic
added
the
enhancement
Codecov Report❌ Patch coverage is
📢 Thoughts on this report? Let us know! |
antekresic
force-pushed
the
antekresic/compaction_poc
branch
from
June 3, 2026 11:20
ff03edc to
2dce6bf
Compare
antekresic
marked this pull request as ready for review
June 3, 2026 13:42
|
@svenklemm, @kpan2034: please review this pull request.
|
svenklemm
reviewed
Jun 16, 2026
svenklemm
reviewed
Jun 16, 2026
svenklemm
reviewed
Jun 16, 2026
antekresic
force-pushed
the
antekresic/compaction_poc
branch
from
June 16, 2026 10:36
2dce6bf to
3a9fb1a
Compare
antekresic
force-pushed
the
antekresic/compaction_poc
branch
from
June 19, 2026 08:08
3a9fb1a to
1d729b9
Compare
akuzm
reviewed
Jun 19, 2026
akuzm
reviewed
Jun 19, 2026
akuzm
approved these changes
Jun 19, 2026
akuzm
left a comment
akuzm
left a comment
Member
There was a problem hiding this comment.
Looks much more straighforward with the new sparse index type.
antekresic
force-pushed
the
antekresic/compaction_poc
branch
2 times, most recently
from
June 23, 2026 09:58
d2e7369 to
71fd91f
Compare
antekresic
enabled auto-merge (rebase)
June 23, 2026 09:58
antekresic
force-pushed
the
antekresic/compaction_poc
branch
from
June 23, 2026 10:00
71fd91f to
d969a2b
Compare
This new function will compact the chunk by looking for overlapping batches and combining them together in order to produce globally ordered chunks. This change is a first step towards supporting direct compress in production workloads.
antekresic
force-pushed
the
antekresic/compaction_poc
branch
from
June 23, 2026 10:08
d969a2b to
e21fd8b
Compare
Closed
Merged
Merged
8 tasks
surister
pushed a commit
that referenced
this pull request
Jul 28, 2026
## 2.29.0 (2026-07-28) This release contains performance improvements and bug fixes since the 2.28.3 release. We recommend that you upgrade at the next available opportunity. **Release Highlights** * **Chunk exclusion for DML operations** drastically improves the performance of `UPDATE` and `DELETE` statements on hypertables. By acquiring exclusive locks only on the specific chunks being modified rather than the entire hypertable, this enhancement eliminates massive lock contention and keeps high-concurrency workloads running smoothly without unnecessary slowdowns. * Intelligent **row-by-row decompression** enables the query planner to decompress data row-by-row rather than in large batches when an operation prioritizes a fast initial response (such as queries with `LIMIT` clauses). This dramatically reduces memory overhead and query latency, ensuring lightning-fast performance when you only need to retrieve a small subset of records from your compressed hypertables. **Important: PostgreSQL 15 Support Removed** TimescaleDB 2.29.0 removes support for PostgreSQL 15. This release supports PostgreSQL 16, 17, and 18. If you are still running PostgreSQL 15, upgrade PostgreSQL before upgrading to TimescaleDB 2.29.0. **Backward-Incompatible Changes** * [#10041](#10041) Remove support for PostgreSQL 15 **Features** * [#9315](#9315) Speed up `DML` operations on hypertables by using the optimized TimescaleDB hypertable expansion code instead of the generic PostgreSQL inheritance hierarchy expansion * [#9534](#9534) Speed up expression evaluation in the columnar pipeline by caching common subexpressions * [#9684](#9684) Add `_timescaledb_functions.decompress_batch()` SQL function * [#9732](#9732) Speed up some queries with small `LIMIT` by switching to row-by-row query execution pipeline * [#9917](#9917) Decompress less data in `DML` on compressed hypertables by accounting for prepared statement parameters * [#9957](#9957) Add `compact_chunk()` function * [#10048](#10048) Support concurrent refresh policies on hierarchical continuous aggregates * [#10081](#10081) Add `samplerate` argument to `_timescaledb_functions.estimate_uncompressed_size()` * [#10100](#10100) Skip classifying compressed relations to speed up planning * [#10118](#10118) Don't track compressed relations as separate chunk * [#10119](#10119) Reduce memory usage of `INSERT` queries using direct compress and spanning multiple chunks * [#10163](#10163) Add a compaction policy for unordered chunks * [#10204](#10204) Don't create separate hypertable catalog entry for hypertables with compression * [#10217](#10217) Initial placeholder version of granular refresh API * [#10225](#10225) Add `config_merge` parameter to `alter_job()` for merging `jsonb` into the existing job configuration * [#10226](#10226) Add `recompress_unordered` columnstore policy option * [#10231](#10231) Use `regclass` for storing relation reference in chunk table * [#10237](#10237) Add helper functions for decoding hypertable status * [#10240](#10240) Add the `tsdb.direct_compress` storage parameter that allows enabling direct compress for a given hypertable independent of global settings * [#10266](#10266) Add `max_batches` to `compact_chunk()` * [#10299](#10299) Add `continuous_aggs_tenant_tracking` and `hypertable_cagg_settings` catalogs **Bugfixes** * [#10013](#10013) Make ownership error messages on continuous aggregates consistent * [#10052](#10052) Result of `MIN` / `MAX` aggregate functions in columnar aggregation pipeline possibly inconsistent with plain PostgreSQL result * [#10071](#10071) Prune the real-time branch of hierarchical continuous aggregates at any nesting depth * [#10143](#10143) Fix division by zero when planning `time_bucket` with zero width * [#10199](#10199) Fix `initial_start` handling in `build_job_info` * [#10213](#10213) Cache sort pathkeys per hypertable * [#10221](#10221) Fix incremental refresh skipping the last bucket * [#10278](#10278) Drop `job_errors` view in `bgw_job_stat_history` migration * [#10280](#10280) `RETURNING` clause returned no rows for `INSERT` using direct compress * [#10281](#10281) Disable direct compress when the destination table has an exclusion constraint so the constraint is still enforced * [#10282](#10282) Only count directly compressed rows toward the command tag when the `INSERT` sets it * [#10286](#10286) Propagate `VACUUM` on a chunk to the compressed relation when running on the chunk directly * [#10302](#10302) Fix useless-join removal and self-join elimination for hypertables * [#10313](#10313) Allow running `ALTER EXTENSION timescaledb UPDATE` inside a transaction block * [#10315](#10315) Fix overlap detection with running max * [#10324](#10324) Fix stale index entries after `rebuild_sparse_index()` on compressed chunks **GUCs** * `timescaledb.enable_hypertable_expansion_for_dml`: allow using the optimized TimescaleDB hypertable expansion code for `UPDATE` and `DELETE` instead of the generic PostgreSQL inheritance hierarchy expansion. On by default. **Thanks** * @FrancescEthon and @ManuelEthon for reporting the issue * @h0rn3t for reporting a problem with `VACUUM` not propagating to the compressed relation * @MaximeEthon for reporting an issue with prepared statement parameters in DML decompression * @proddata for reporting a problem when upgrading from 2.15.3 to 2.28.2 * @tureba for reporting and fixing stale sparse-index entries after rebuild * @viniciusrsouza for reporting an issue with hierarchical continuous aggregates
svenklemm
pushed a commit
that referenced
this pull request
Jul 28, 2026
This release contains performance improvements and bug fixes since the 2.28.3 release. We recommend that you upgrade at the next available opportunity. **Release Highlights** * **Chunk exclusion for DML operations** drastically improves the performance of `UPDATE` and `DELETE` statements on hypertables. By acquiring exclusive locks only on the specific chunks being modified rather than the entire hypertable, this enhancement eliminates massive lock contention and keeps high-concurrency workloads running smoothly without unnecessary slowdowns. * Intelligent **row-by-row decompression** enables the query planner to decompress data row-by-row rather than in large batches when an operation prioritizes a fast initial response (such as queries with `LIMIT` clauses). This dramatically reduces memory overhead and query latency, ensuring lightning-fast performance when you only need to retrieve a small subset of records from your compressed hypertables. **Important: PostgreSQL 15 Support Removed** TimescaleDB 2.29.0 removes support for PostgreSQL 15. This release supports PostgreSQL 16, 17, and 18. If you are still running PostgreSQL 15, upgrade PostgreSQL before upgrading to TimescaleDB 2.29.0. **Backward-Incompatible Changes** * [#10041](#10041) Remove support for PostgreSQL 15 **Features** * [#9315](#9315) Speed up `DML` operations on hypertables by using the optimized TimescaleDB hypertable expansion code instead of the generic PostgreSQL inheritance hierarchy expansion * [#9534](#9534) Speed up expression evaluation in the columnar pipeline by caching common subexpressions * [#9684](#9684) Add `_timescaledb_functions.decompress_batch()` SQL function * [#9732](#9732) Speed up some queries with small `LIMIT` by switching to row-by-row query execution pipeline * [#9917](#9917) Decompress less data in `DML` on compressed hypertables by accounting for prepared statement parameters * [#9957](#9957) Add `compact_chunk()` function * [#10048](#10048) Support concurrent refresh policies on hierarchical continuous aggregates * [#10081](#10081) Add `samplerate` argument to `_timescaledb_functions.estimate_uncompressed_size()` * [#10100](#10100) Skip classifying compressed relations to speed up planning * [#10118](#10118) Don't track compressed relations as separate chunk * [#10119](#10119) Reduce memory usage of `INSERT` queries using direct compress and spanning multiple chunks * [#10163](#10163) Add a compaction policy for unordered chunks * [#10204](#10204) Don't create separate hypertable catalog entry for hypertables with compression * [#10217](#10217) Initial placeholder version of granular refresh API * [#10225](#10225) Add `config_merge` parameter to `alter_job()` for merging `jsonb` into the existing job configuration * [#10226](#10226) Add `recompress_unordered` columnstore policy option * [#10231](#10231) Use `regclass` for storing relation reference in chunk table * [#10237](#10237) Add helper functions for decoding hypertable status * [#10240](#10240) Add the `tsdb.direct_compress` storage parameter that allows enabling direct compress for a given hypertable independent of global settings * [#10266](#10266) Add `max_batches` to `compact_chunk()` * [#10299](#10299) Add `continuous_aggs_tenant_tracking` and `hypertable_cagg_settings` catalogs **Bugfixes** * [#10013](#10013) Make ownership error messages on continuous aggregates consistent * [#10052](#10052) Result of `MIN` / `MAX` aggregate functions in columnar aggregation pipeline possibly inconsistent with plain PostgreSQL result * [#10071](#10071) Prune the real-time branch of hierarchical continuous aggregates at any nesting depth * [#10143](#10143) Fix division by zero when planning `time_bucket` with zero width * [#10199](#10199) Fix `initial_start` handling in `build_job_info` * [#10213](#10213) Cache sort pathkeys per hypertable * [#10221](#10221) Fix incremental refresh skipping the last bucket * [#10278](#10278) Drop `job_errors` view in `bgw_job_stat_history` migration * [#10280](#10280) `RETURNING` clause returned no rows for `INSERT` using direct compress * [#10281](#10281) Disable direct compress when the destination table has an exclusion constraint so the constraint is still enforced * [#10282](#10282) Only count directly compressed rows toward the command tag when the `INSERT` sets it * [#10286](#10286) Propagate `VACUUM` on a chunk to the compressed relation when running on the chunk directly * [#10302](#10302) Fix useless-join removal and self-join elimination for hypertables * [#10313](#10313) Allow running `ALTER EXTENSION timescaledb UPDATE` inside a transaction block * [#10315](#10315) Fix overlap detection with running max * [#10324](#10324) Fix stale index entries after `rebuild_sparse_index()` on compressed chunks **GUCs** * `timescaledb.enable_hypertable_expansion_for_dml`: allow using the optimized TimescaleDB hypertable expansion code for `UPDATE` and `DELETE` instead of the generic PostgreSQL inheritance hierarchy expansion. On by default. **Thanks** * @FrancescEthon and @ManuelEthon for reporting an issue with incremental refresh skipping the last bucket * @h0rn3t for reporting a problem with `VACUUM` not propagating to the * @igor2x for reporting an issue with locking during DML statements on hypertables compressed relation * @MaximeEthon for reporting an issue with prepared statement parameters in DML decompression * @proddata for reporting a problem when upgrading from 2.15.3 to 2.28.2 * @tureba for reporting and fixing stale sparse-index entries after rebuild * @viniciusrsouza for reporting an issue with hierarchical continuous aggregates
surister
pushed a commit
to surister/timescaledb
that referenced
this pull request
Jul 28, 2026
This release contains performance improvements and bug fixes since the 2.28.3 release. We recommend that you upgrade at the next available opportunity. **Release Highlights** * **Chunk exclusion for DML operations** drastically improves the performance of `UPDATE` and `DELETE` statements on hypertables. By acquiring exclusive locks only on the specific chunks being modified rather than the entire hypertable, this enhancement eliminates massive lock contention and keeps high-concurrency workloads running smoothly without unnecessary slowdowns. * Intelligent **row-by-row decompression** enables the query planner to decompress data row-by-row rather than in large batches when an operation prioritizes a fast initial response (such as queries with `LIMIT` clauses). This dramatically reduces memory overhead and query latency, ensuring lightning-fast performance when you only need to retrieve a small subset of records from your compressed hypertables. **Important: PostgreSQL 15 Support Removed** TimescaleDB 2.29.0 removes support for PostgreSQL 15. This release supports PostgreSQL 16, 17, and 18. If you are still running PostgreSQL 15, upgrade PostgreSQL before upgrading to TimescaleDB 2.29.0. **Backward-Incompatible Changes** * [timescale#10041](timescale#10041) Remove support for PostgreSQL 15 **Features** * [timescale#9315](timescale#9315) Speed up `DML` operations on hypertables by using the optimized TimescaleDB hypertable expansion code instead of the generic PostgreSQL inheritance hierarchy expansion * [timescale#9534](timescale#9534) Speed up expression evaluation in the columnar pipeline by caching common subexpressions * [timescale#9684](timescale#9684) Add `_timescaledb_functions.decompress_batch()` SQL function * [timescale#9732](timescale#9732) Speed up some queries with small `LIMIT` by switching to row-by-row query execution pipeline * [timescale#9917](timescale#9917) Decompress less data in `DML` on compressed hypertables by accounting for prepared statement parameters * [timescale#9957](timescale#9957) Add `compact_chunk()` function * [timescale#10048](timescale#10048) Support concurrent refresh policies on hierarchical continuous aggregates * [timescale#10081](timescale#10081) Add `samplerate` argument to `_timescaledb_functions.estimate_uncompressed_size()` * [timescale#10100](timescale#10100) Skip classifying compressed relations to speed up planning * [timescale#10118](timescale#10118) Don't track compressed relations as separate chunk * [timescale#10119](timescale#10119) Reduce memory usage of `INSERT` queries using direct compress and spanning multiple chunks * [timescale#10163](timescale#10163) Add a compaction policy for unordered chunks * [timescale#10204](timescale#10204) Don't create separate hypertable catalog entry for hypertables with compression * [timescale#10217](timescale#10217) Initial placeholder version of granular refresh API * [timescale#10225](timescale#10225) Add `config_merge` parameter to `alter_job()` for merging `jsonb` into the existing job configuration * [timescale#10226](timescale#10226) Add `recompress_unordered` columnstore policy option * [timescale#10231](timescale#10231) Use `regclass` for storing relation reference in chunk table * [timescale#10237](timescale#10237) Add helper functions for decoding hypertable status * [timescale#10240](timescale#10240) Add the `tsdb.direct_compress` storage parameter that allows enabling direct compress for a given hypertable independent of global settings * [timescale#10266](timescale#10266) Add `max_batches` to `compact_chunk()` * [timescale#10299](timescale#10299) Add `continuous_aggs_tenant_tracking` and `hypertable_cagg_settings` catalogs **Bugfixes** * [timescale#10013](timescale#10013) Make ownership error messages on continuous aggregates consistent * [timescale#10052](timescale#10052) Result of `MIN` / `MAX` aggregate functions in columnar aggregation pipeline possibly inconsistent with plain PostgreSQL result * [timescale#10071](timescale#10071) Prune the real-time branch of hierarchical continuous aggregates at any nesting depth * [timescale#10143](timescale#10143) Fix division by zero when planning `time_bucket` with zero width * [timescale#10199](timescale#10199) Fix `initial_start` handling in `build_job_info` * [timescale#10213](timescale#10213) Cache sort pathkeys per hypertable * [timescale#10221](timescale#10221) Fix incremental refresh skipping the last bucket * [timescale#10278](timescale#10278) Drop `job_errors` view in `bgw_job_stat_history` migration * [timescale#10280](timescale#10280) `RETURNING` clause returned no rows for `INSERT` using direct compress * [timescale#10281](timescale#10281) Disable direct compress when the destination table has an exclusion constraint so the constraint is still enforced * [timescale#10282](timescale#10282) Only count directly compressed rows toward the command tag when the `INSERT` sets it * [timescale#10286](timescale#10286) Propagate `VACUUM` on a chunk to the compressed relation when running on the chunk directly * [timescale#10302](timescale#10302) Fix useless-join removal and self-join elimination for hypertables * [timescale#10313](timescale#10313) Allow running `ALTER EXTENSION timescaledb UPDATE` inside a transaction block * [timescale#10315](timescale#10315) Fix overlap detection with running max * [timescale#10324](timescale#10324) Fix stale index entries after `rebuild_sparse_index()` on compressed chunks **GUCs** * `timescaledb.enable_hypertable_expansion_for_dml`: allow using the optimized TimescaleDB hypertable expansion code for `UPDATE` and `DELETE` instead of the generic PostgreSQL inheritance hierarchy expansion. On by default. **Thanks** * @FrancescEthon and @ManuelEthon for reporting an issue with incremental refresh skipping the last bucket * @h0rn3t for reporting a problem with `VACUUM` not propagating to the * @igor2x for reporting an issue with locking during DML statements on hypertables compressed relation * @MaximeEthon for reporting an issue with prepared statement parameters in DML decompression * @proddata for reporting a problem when upgrading from 2.15.3 to 2.28.2 * @tureba for reporting and fixing stale sparse-index entries after rebuild * @viniciusrsouza for reporting an issue with hierarchical continuous aggregates
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
4 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
This new function will compact the chunk by looking for overlapping batches and combining them together in order to produce globally ordered chunks. This is change is a first step towards supporting direct compress in production workloads.