From 0c562d2aa728424c397badc2bfd02126ee983c13 Mon Sep 17 00:00:00 2001 From: Dilip Kumar Date: Thu, 14 Aug 2025 07:19:49 +0000 Subject: [PATCH v5] Fix DROP SUBSCRIPTION deadlock with new database creation DROP SUBSCRIPTION previously acquired an AccessExclusiveLock on the pg_subscription catalog to prevent the replication launcher from starting a new worker. However, this caused a deadlock. New database creation also need to acquire an AccessShareLock on this same catalog during their initialization phase. This created a lock conflict where DROP SUBSCRIPTION would block this, while simultaneously waiting for a connection to complete for dropping the replication slot. This commit resolves the deadlock by having DROP SUBSCRIPTION acquire a less restrictive RowExclusiveLock on the catalog instead. To address the original concern of orphaned workers, a new check is implemented. The replication worker now takes a shared object lock on the subscription itself. If a worker starts for a subscription that no longer exists, it immediately detects this condition and exits. This ensures that no orphan workers are created without the need for an overly broad AccessExclusiveLock on the system catalog. --- src/backend/commands/subscriptioncmds.c | 8 +++++--- src/backend/replication/logical/worker.c | 7 +++++++ src/test/regress/expected/subscription.out | 13 +++++++++++++ src/test/regress/sql/subscription.sql | 13 +++++++++++++ 4 files changed, 38 insertions(+), 3 deletions(-) diff --git a/src/backend/commands/subscriptioncmds.c b/src/backend/commands/subscriptioncmds.c index 40356e97ed2..28ab74ece93 100644 --- a/src/backend/commands/subscriptioncmds.c +++ b/src/backend/commands/subscriptioncmds.c @@ -844,10 +844,12 @@ DropSubscription(DropSubscriptionStmt *stmt, bool isTopLevel) Form_pg_subscription form; /* - * Lock pg_subscription with AccessExclusiveLock to ensure that the - * launcher doesn't restart new worker during dropping the subscription + * The launcher may concurrently start a new worker for this subscription. + * During initialization, the worker checks for subscription validity and + * exits if the subscription has already been dropped. See + * InitializeLogRepWorker. */ - rel = table_open(SubscriptionRelationId, AccessExclusiveLock); + rel = table_open(SubscriptionRelationId, RowExclusiveLock); tup = SearchSysCache2(SUBSCRIPTIONNAME, MyDatabaseId, CStringGetDatum(stmt->subname)); diff --git a/src/backend/replication/logical/worker.c b/src/backend/replication/logical/worker.c index a0e64301a94..b41738627ca 100644 --- a/src/backend/replication/logical/worker.c +++ b/src/backend/replication/logical/worker.c @@ -2070,6 +2070,13 @@ ApplyWorkerMain(Datum main_arg) StartTransactionCommand(); oldctx = MemoryContextSwitchTo(ApplyContext); + /* + * Lock the subscription to prevent it from being concurrently dropped, + * then re-verify its existence. After the initialization, the worker will + * be terminated gracefully if the subscription is dropped. + */ + LockSharedObject(SubscriptionRelationId, MyLogicalRepWorker->subid, 0, + AccessShareLock); MySubscription = GetSubscription(MyLogicalRepWorker->subid, true); if (!MySubscription) { diff --git a/src/test/regress/expected/subscription.out b/src/test/regress/expected/subscription.out index 67cb15f4f58..b59e6e1af38 100644 --- a/src/test/regress/expected/subscription.out +++ b/src/test/regress/expected/subscription.out @@ -173,6 +173,19 @@ DROP SUBSCRIPTION IF EXISTS regress_testsub; NOTICE: subscription "regress_testsub" does not exist, skipping DROP SUBSCRIPTION regress_testsub; -- fail ERROR: subscription "regress_testsub" does not exist +-- this test verifies that DROP SUBSCRIPTION returns an immediate error +-- and does not hang when it makes connection to a recently created database +-- whose caches are not yet initialized. +CREATE DATABASE regression_db; +CREATE SUBSCRIPTION regress_testsub CONNECTION 'dbname=regression_db' PUBLICATION testpub WITH (connect=false); +WARNING: tables were not subscribed, you will have to run ALTER SUBSCRIPTION ... REFRESH PUBLICATION to subscribe the tables +DROP SUBSCRIPTION regress_testsub; +ERROR: could not drop the replication slot "regress_testsub" on publisher +DETAIL: The error was: ERROR: replication slot "regress_testsub" does not exist +-- now set the slot_name to NONE and DROP SUBSCRIPTION, it should work +ALTER SUBSCRIPTION regress_testsub SET (slot_name = NONE); +DROP SUBSCRIPTION regress_testsub; +DROP DATABASE regression_db; RESET SESSION AUTHORIZATION; DROP ROLE regress_subscription_user; DROP ROLE regress_subscription_user2; diff --git a/src/test/regress/sql/subscription.sql b/src/test/regress/sql/subscription.sql index f0d38ad19bb..44cc31a113e 100644 --- a/src/test/regress/sql/subscription.sql +++ b/src/test/regress/sql/subscription.sql @@ -129,6 +129,19 @@ COMMIT; DROP SUBSCRIPTION IF EXISTS regress_testsub; DROP SUBSCRIPTION regress_testsub; -- fail +-- this test verifies that DROP SUBSCRIPTION returns an immediate error +-- and does not hang when it makes connection to a recently created database +-- whose caches are not yet initialized. +CREATE DATABASE regression_db; +CREATE SUBSCRIPTION regress_testsub CONNECTION 'dbname=regression_db' PUBLICATION testpub WITH (connect=false); + +DROP SUBSCRIPTION regress_testsub; + +-- now set the slot_name to NONE and DROP SUBSCRIPTION, it should work +ALTER SUBSCRIPTION regress_testsub SET (slot_name = NONE); +DROP SUBSCRIPTION regress_testsub; +DROP DATABASE regression_db; + RESET SESSION AUTHORIZATION; DROP ROLE regress_subscription_user; DROP ROLE regress_subscription_user2; -- 2.51.0.rc1.163.g2494970778-goog