Import gp_relsizes_stats extention from Greenplum

[Vlasdislav]

What does this PR do?

This PR ports the gp_relsizes_stats extension from the standalone open-gpdb/gp_relsizes_stats repository into the Cloudberry monorepo under gpcontrib/ and adapts it to Cloudberry (PostgreSQL 14 base).

The extension collects and stores statistics about table and file sizes on coordinator and segment hosts. It supports automatic collection via a Background Worker and manual one-shot collection via relsizes_stats_schema.relsizes_collect_stats_once().

Build / integration changes:

• Makefile — added dual-mode build support: USE_PGXS=1 for standalone builds (original behavior) and in-tree build via $(top_builddir)/src/Makefile.global + contrib-global.mk for building inside the Cloudberry source tree.
• gpcontrib/Makefile — added gp_relsizes_stats to recurse_targets for both release and debug (enable_debug_extensions=yes) build configurations.
• README.md — updated terminology (master → coordinator, GPDB → Cloudberry) and installation instructions for in-tree build.

SQL adaptation for Cloudberry / PG14 (sql/gp_relsizes_stats--1.3.sql and sql/gp_relsizes_stats--1.1--1.2.sql):

• Rewrote the table_files view to use a recursive WITH RECURSIVE walk over pg_inherits instead of the GP6-only pg_partition /
  pg_partition_rule catalogs, which do not exist in Cloudberry.
• Switched the join with segment_file_sizes from INNER to LEFT JOIN (with COALESCE(size, 0) / COALESCE(mtime, 0)): on Cloudberry/PG14 the partition root and intermediate partitioned tables have relfilenode = 0 and therefore have no row in segment_file_sizes; we still want them surfaced with size = 0.
• Removed OIDS=FALSE from segment_file_sizes (unsupported since PG12).
• Removed the TRUNCATE TABLE segment_file_sizes / TRUNCATE TABLE table_sizes_history calls that were executed during extension install — re-installing the extension no longer wipes previously collected stats.
• Dropped EXECUTE ON MASTER from relsizes_collect_stats_once() (the notion of MASTER is replaced by COORDINATOR in Cloudberry; the default dispatch is sufficient).

Type of Change

• Bug fix (non-breaking change)
• New feature (non-breaking change)
• Breaking change (fix or feature with breaking changes)
• Documentation update

Breaking Changes

Not applicable — this is a new extension; nothing existing in the tree changes behavior.

Impact

Performance:
No impact on existing functionality. The background worker is disabled by default (gp_relsizes_stats.enabled = false). When enabled, naptimes between databases / files are tunable to spread the load.

User-facing changes:
New extension gp_relsizes_stats available under gpcontrib. Must be added to shared_preload_libraries to enable the background worker. Manual collection via SELECT relsizes_stats_schema.relsizes_collect_stats_once(); does not require preload.

Dependencies:
None beyond standard PostgreSQL / Cloudberry server headers (bgworker, shmem, cdbvars, pg_appendonly).

Additional Context

Upstream repository: https://github.com/open-gpdb/gp_relsizes_stats

[github-actions]

Hi, @Vlasdislav welcome!🎊 Thanks for taking the effort to make our project better! 🙌 Keep making such awesome contributions!

[Copilot]

Pull request overview

This PR ports the gp_relsizes_stats extension into the Cloudberry monorepo under gpcontrib/, integrating it into the in-tree build and adding regression tests.

Changes:

• Adds the new gp_relsizes_stats extension (C code, SQL install/upgrade scripts, control file, docs).
• Integrates the extension into gpcontrib/Makefile recursion targets and adds a dual-mode (PGXS vs in-tree) Makefile.
• Adds regression tests and expected outputs for functionality and privilege behavior.

Reviewed changes

Copilot reviewed 16 out of 16 changed files in this pull request and generated 15 comments.

Show a summary per file

File	Description
gpcontrib/Makefile	Adds gp_relsizes_stats to the gpcontrib build recurse targets.
gpcontrib/gp_relsizes_stats/Makefile	Builds/installs the extension in-tree or via PGXS; wires up regression tests.
gpcontrib/gp_relsizes_stats/src/gp_relsizes_stats.c	Implements background worker orchestration and filesystem stats collection.
gpcontrib/gp_relsizes_stats/sql/gp_relsizes_stats--1.3.sql	Creates schema/tables/views and declares C functions.
gpcontrib/gp_relsizes_stats/sql/gp_relsizes_stats--1.2--1.3.sql	Upgrade script (grants).
gpcontrib/gp_relsizes_stats/sql/gp_relsizes_stats--1.1--1.2.sql	Upgrade script (view changes).
gpcontrib/gp_relsizes_stats/sql/gp_relsizes_stats--1.0--1.1.sql	Upgrade script (function signature change).
gpcontrib/gp_relsizes_stats/gp_relsizes_stats.control	Registers extension metadata (version, module path, trusted flag).
gpcontrib/gp_relsizes_stats/test/sql/grants.sql	Adds privileges/regression coverage for creator + granting to another role.
gpcontrib/gp_relsizes_stats/test/sql/gp_relsizes_stats.sql	Adds functional regression coverage for stats collection and size reporting.
gpcontrib/gp_relsizes_stats/test/expected/grants.out	Expected output for grants.sql.
gpcontrib/gp_relsizes_stats/test/expected/gp_relsizes_stats.out	Expected output for gp_relsizes_stats.sql.
gpcontrib/gp_relsizes_stats/README.md	Extension documentation.
gpcontrib/gp_relsizes_stats/LICENCE	Bundled Apache 2.0 license text for the extension.
gpcontrib/gp_relsizes_stats/.gitignore	Ignores local build artifacts in the extension directory.
gpcontrib/gp_relsizes_stats/.clang-format	Local formatting configuration for the extension sources.

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[tuhaihe]

Hi @Vlasdislav thanks for your work. The following are my comments for your reference:

1. Import the commit history to the gpcontrib/gp_relsizes_stats from the start for a better rebase:

git filter-repo --force --to-subdirectory-filter gpcontrib/gp_relsizes_stats

2. Rewrite some commits for a better clean & lean commit history

3. ASF license compliance. See more comments in the code.

[tuhaihe]

Hi @Vladislav,

Thank you for your work. Here are my comments for your consideration:

1. Please re-import the commit history into the gpcontrib/*** directory instead of placing it in the root directory -- from the first commit, though moving the files to gpcontrib at last. This change will help prevent conflicts with the top source code files in Cloudberry when rebasing the PR.
2. It would be beneficial to clean up the commit history.

You can refer to this PR for guidance: #1629.

[Vlasdislav]

Hi @Vladislav,

Thank you for your work. Here are my comments for your consideration:

1. Please re-import the commit history into the gpcontrib/*** directory instead of placing it in the root directory -- from the first commit, though moving the files to gpcontrib at last. This change will help prevent conflicts with the top source code files in Cloudberry when rebasing the PR.
2. It would be beneficial to clean up the commit history.

You can refer to this PR for guidance: #1629.

Hi, @tuhaihe, See if I understood you correctly.

[leborchuk]

Checked - all looks good for me, much more better than original version

[leborchuk]

See regression in tests, check https://github.com/apache/cloudberry/actions/runs/26833894082/job/79174723029?pr=1757 - some issues with history stat and partition info

[Vlasdislav]

See regression in tests, check https://github.com/apache/cloudberry/actions/runs/26833894082/job/79174723029?pr=1757 - some issues with history stat and partition info

If I understood your message correctly, then I looked at the error and the problem is not in the code (local it's OK), but in the configuration of running the CI test. There may be a fix in this commit: f07716d.

[tuhaihe]

Overall, LGTM now from the non-tech side, like the ASF-compliance related. Would like to have your technical comments on this PR, cc @my-ship-it @Smyatkin-Maxim, @yjhjstz

[Smyatkin-Maxim]

cc @my-ship-it @Smyatkin-Maxim, @yjhjstz

Will take a look

[my-ship-it]

LGTM, thanks

[Smyatkin-Maxim]

LGTM. I think we have to squash it to just two commits: @AndrewOvvv `s initial commit to keep authorship and the other one with @Vlasdislav changes.

[tuhaihe]

Hi @Vlasdislav we are unable to complete the rebase and merge process due to certain conflicts. Perhaps you could try rebasing on your local machine.

[Vlasdislav]

Hi @Vlasdislav we are unable to complete the rebase and merge process due to certain conflicts. Perhaps you could try rebasing on your local machine.

Hi, @tuhaihe should be OK now