From d605cde3b7c8a56700262f553c17172b627fd413 Mon Sep 17 00:00:00 2001 From: Naisila Puka <37271756+naisila@users.noreply.github.com> Date: Thu, 1 Sep 2022 11:56:31 +0300 Subject: [PATCH] Use RelationGetPrimaryKeyIndex for citus catalog tables (#6262) pg_dist_node and pg_dist_colocation have a primary key index, not a replica identity index. Citus catalog tables are created in public schema, which has replica identity index by default as primary key index. Later the citus catalog tables are moved to pg_catalog schema. During pg_upgrade, all tables are recreated, and given that pg_dist_colocation is found in pg_catalog schema, it is recreated in that schema, and when it is recreated it doesn't have a replica identity index, because catalog tables have no replica identity. Further action: Do we even need to acquire this lock on the primary key index? Postgres doesn't acquire such locks on indexes before deleting catalog tuples. Also, catalog tuples don't have replica identities by definition. --- src/backend/distributed/metadata/node_metadata.c | 11 +++++++++-- src/backend/distributed/utils/colocation_utils.c | 11 +++++++++-- 2 files changed, 18 insertions(+), 4 deletions(-) diff --git a/src/backend/distributed/metadata/node_metadata.c b/src/backend/distributed/metadata/node_metadata.c index 6da08c02e..f1814c1eb 100644 --- a/src/backend/distributed/metadata/node_metadata.c +++ b/src/backend/distributed/metadata/node_metadata.c @@ -2026,9 +2026,16 @@ DeleteNodeRow(char *nodeName, int32 nodePort) /* * simple_heap_delete() expects that the caller has at least an - * AccessShareLock on replica identity index. + * AccessShareLock on primary key index. + * + * XXX: This does not seem required, do we really need to acquire this lock? + * Postgres doesn't acquire such locks on indexes before deleting catalog tuples. + * Linking here the reasons we added this lock acquirement: + * https://github.com/citusdata/citus/pull/2851#discussion_r306569462 + * https://github.com/citusdata/citus/pull/2855#discussion_r313628554 + * https://github.com/citusdata/citus/issues/1890 */ - Relation replicaIndex = index_open(RelationGetReplicaIndex(pgDistNode), + Relation replicaIndex = index_open(RelationGetPrimaryKeyIndex(pgDistNode), AccessShareLock); ScanKeyInit(&scanKey[0], Anum_pg_dist_node_nodename, diff --git a/src/backend/distributed/utils/colocation_utils.c b/src/backend/distributed/utils/colocation_utils.c index fa698f88a..50722dd9a 100644 --- a/src/backend/distributed/utils/colocation_utils.c +++ b/src/backend/distributed/utils/colocation_utils.c @@ -1235,10 +1235,17 @@ DeleteColocationGroup(uint32 colocationId) { /* * simple_heap_delete() expects that the caller has at least an - * AccessShareLock on replica identity index. + * AccessShareLock on primary key index. + * + * XXX: This does not seem required, do we really need to acquire this lock? + * Postgres doesn't acquire such locks on indexes before deleting catalog tuples. + * Linking here the reasons we added this lock acquirement: + * https://github.com/citusdata/citus/pull/2851#discussion_r306569462 + * https://github.com/citusdata/citus/pull/2855#discussion_r313628554 + * https://github.com/citusdata/citus/issues/1890 */ Relation replicaIndex = - index_open(RelationGetReplicaIndex(pgDistColocation), + index_open(RelationGetPrimaryKeyIndex(pgDistColocation), AccessShareLock); simple_heap_delete(pgDistColocation, &(heapTuple->t_self));