Releases · microsoft/pg_durable

Release list

Release v0.2.4-rc1 Pre-release

Pre-release

github-actions released this 02 Jul 22:52

v0.2.4-rc1

05cd2b6

Release candidate for v0.2.4. This is a pre-release for validation only — not a final release. The notes below describe the full v0.2.4 changes.

Provider-line note: v0.2.4 stays in the duroxide-pg provider compatibility line, so the upgrade source is v0.2.3 (sql/pg_durable--0.2.3--0.2.4.sql).

Added

Instance retention/pruning (#265): terminal instances are now pruned by a hard cap and a retention window, bounding unbounded growth of df.instances.
df.list_instances() pagination & filtering (#278): added a label_filter, keyset pagination (after_cursor/next_cursor), and created_at/updated_at timestamps to the result set.

Changed

df.wait_for_schedule() cron timing: the next cron tick is now computed at execution time using duroxide's deterministic clock (ctx.utc_now()) inside the execute_function_graph orchestration, instead of being pre-computed at df.start() time. This makes recurring @> schedules and any start-to-execution delay target the correct upcoming tick (#130).

⚠️ Replay-breaking for in-flight wait_for_schedule instances. This change adds a recorded utc_now() decision before the WAIT_SCHEDULE timer, altering the orchestration's history sequence. Any durable function that was started under a <= 0.2.3 binary and is mid-wait_for_schedule (parked on its timer) when this .so is loaded will fail with a duroxide nondeterminism error on replay, because its recorded history no longer matches the new code. Drain or allow such in-flight wait_for_schedule instances to complete before upgrading. Instances that are not currently inside a wait_for_schedule node are unaffected. We accepted this break (rather than introducing orchestration versioning) given the early pre-1.0 stage of the project.
Instance/node ID collision hardening (#129): df.start() now reserves IDs with INSERT ... ON CONFLICT DO NOTHING RETURNING id and re-rolls the random 8-hex value on collision — instances arbitrate on the df.instances primary key (id), nodes on the new composite PRIMARY KEY (instance_id, id) — replacing the previous SELECT EXISTS pre-check. Doing the conflict check at the index level (rather than a pre-check SELECT) closes a TOCTOU window and, for instances, an RLS blind spot where the pre-check could not see another role's rows. df.nodes now uses the composite PRIMARY KEY (instance_id, id) instead of a global single-column key, so the random 8-hex node ID is no longer the sole cross-instance collision guard. The update-node-status activity now scopes its df.nodes update by instance_id (a required activity-input field) and asserts it affects exactly one row. IDs stay VARCHAR(8) HEX; the 0.2.3 → 0.2.4 upgrade restructures the df.nodes keys in place (#238).
- Breaking for in-flight work: the new activity-input shape changes the string duroxide records in orchestration history, and duroxide validates activity inputs by exact equality on replay, so any instance left in flight across the 0.2.3 → 0.2.4 binary upgrade cannot complete. Drain or cancel in-flight instances before deploying 0.2.4. The in-place df.nodes key restructure also takes an ACCESS EXCLUSIVE lock whose duration scales with table size — run the upgrade in a maintenance window. See the #129 section of docs/upgrade-testing.md for the full drain-before-upgrade contract.
df.grant_usage() / df.revoke_usage(): dropped the explicit per-function EXECUTE allowlist. Schema USAGE on df is the real access gate for ordinary df.* functions, so the helpers now grant/revoke schema USAGE, the table privileges, and EXECUTE only on the sensitive functions (df.http, df.grant_usage, df.revoke_usage). Function signatures are unchanged and existing privileges are unaffected (#242).
df.list_instances() page-size cap is now a loud error (#146): df.list_instances() previously truncated limit_count silently to a fixed ceiling of 10000. It now raises an error when limit_count exceeds the new pg_durable.list_instances_max_limit GUC (SUSET context, default 1000, range 1–1000000), so an over-cap request fails fast instead of returning a silently short page that is indistinguishable from "no more rows". Both the basic and paginated overloads enforce the cap; clients needing more rows should lower limit_count or use the paginated overload (after_cursor/next_cursor). A superuser can change the cap at runtime without a restart; by default an ordinary caller cannot.
Renamed df.wait_for_completion() (#164): the function was renamed and hardened against unsafe use. Breaking: callers of the old name must update to the new name.
Node statuses derived from execution lineage (#263, #283): node status is now derived from the durable engine's execution lineage, reconciling the df control-plane with the engine so reported statuses match actual execution.
df.start() fails fast on engine hand-off failure (#282): if the hand-off to the durable engine fails, df.start() now returns an error immediately instead of leaving a stuck instance behind.
Dependencies: bumped reqwest to 0.13.4 (#260) and uuid to 1.23.4 (#273).

Fixed

explain race branches (#276): race (|) branches now render correctly in df.explain() output.
Loop safety (#254): df.loop() now enforces a max-iteration guard and detects malformed loop configuration instead of looping unboundedly or misbehaving.
$name.* expansion cap (#255): a row-count limit (10,000) is now enforced when expanding $name.*, preventing unbounded expansion.
df.http() User-Agent (#270): requests now send a default User-Agent header.
Connection reliability (#251, #252): the client is now recoverable after a connection failure, and epoch/extension polling is isolated onto a dedicated connection pool so it can no longer contend with execution work.
df.list_instances() N+1 (#275): instance-info lookups are now batched, removing an N+1 query pattern.
Indexes (#271): added a created_at index and a composite status index on df.instances to speed up listing and status queries.

Security

SSRF CGNAT range (#253): the 100.64.0.0/10 CGNAT range is now blocked by SSRF protection in df.http().
df.metrics() access (#184): df.metrics() is now gated behind an explicit EXECUTE grant rather than being callable by default.

Removed

df.debug_connection(): removed from the SQL surface as non-security, surface-reduction cleanup (#110). The function returned the worker connection string (postgres://role@host:port/db) with no password or credential, and the worker role is already visible through native PostgreSQL channels (the world-readable pg_durable.worker_role GUC and pg_stat_activity.usename) — so issue #110 is reclassified from a security finding to cleanup. Fresh installs no longer create the function and the 0.2.3 → 0.2.4 upgrade drops it; a binary-compatibility shim retains the underlying C symbol so pre-0.2.4 schemas keep resolving the function until ALTER EXTENSION pg_durable UPDATE runs.

Documentation

Documented SECURITY DEFINER df.start() behavior (#185), corrected documentation examples (#257), and clarified that df.status()/df.result() take an instance_id rather than a label (#167).

Acknowledgements

Thanks to everyone who contributed to this release: @crprashant @iemejia @nanookclaw @pinodeca @Saithej2k @snvtac @tjgreen42.

New Contributors

@snvtac made their first contribution in #262
@nanookclaw made their first contribution in #265
@Saithej2k made their first contribution in #276

Full Changelog: v0.2.3...v0.2.4-rc1

Contributors

iemejia, tjgreen42, and 5 other contributors

Assets 7

Release v0.2.3 Latest

Latest

github-actions released this 17 Jun 22:42

v0.2.3

b4579ae

What's new in v0.2.3

Provider-line note: v0.2.3 stays in the duroxide-pg provider compatibility line started in v0.2.2, so the upgrade source is v0.2.2 (sql/pg_durable--0.2.2--0.2.3.sql).