From 9cd1bd70848341995cd0117432dbaef39576df23 Mon Sep 17 00:00:00 2001 From: Dilip Kumar Date: Thu, 14 Aug 2025 07:12:34 +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 | 12 ++++++++++++ src/test/regress/sql/subscription.sql | 13 +++++++++++++ 4 files changed, 37 insertions(+), 3 deletions(-) diff --git a/src/backend/commands/subscriptioncmds.c b/src/backend/commands/subscriptioncmds.c index b1a2f3f81a2..c09ddbc7a22 100644 --- a/src/backend/commands/subscriptioncmds.c +++ b/src/backend/commands/subscriptioncmds.c @@ -1070,10 +1070,12 @@ DropSubscription(DropSubscriptionStmt *stmt, bool isTopLevel) List *rstates; /* - * 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 d1c7d352363..93a7714922e 100644 --- a/src/backend/replication/logical/worker.c +++ b/src/backend/replication/logical/worker.c @@ -3129,6 +3129,13 @@ 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 85bd97ed8dd..d067c639416 100644 --- a/src/test/regress/expected/subscription.out +++ b/src/test/regress/expected/subscription.out @@ -271,6 +271,18 @@ ALTER SUBSCRIPTION regress_testsub DISABLE; ALTER SUBSCRIPTION regress_testsub SET (slot_name = NONE); DROP SUBSCRIPTION regress_testsub; DROP FUNCTION func; +-- 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 replication slot "regress_testsub" on publisher: 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 c8a2d08e721..b117d30c3e1 100644 --- a/src/test/regress/sql/subscription.sql +++ b/src/test/regress/sql/subscription.sql @@ -209,6 +209,19 @@ ALTER SUBSCRIPTION regress_testsub SET (slot_name = NONE); DROP SUBSCRIPTION regress_testsub; DROP FUNCTION func; +-- 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