Chao Li <[email protected]> wrote:
> I am continuing to test REPACK, and I found another issue.
>
> In check_concurrent_repack_requirements(), if a table has no replica identity
> index, the code falls back to using the primary key if one exists. The
> problem is that a deferrable primary key cannot be used for this purpose. WAL
> generation does not consider a deferrable primary key to be a replica
> identity, so concurrent mode may not receive enough old tuple information to
> replay concurrent changes.
Thanks for finding it, this is certainly a problem.
I'm just thinking if it's worth a separate error message.
RelationGetIndexList() just ignores the deferrable PK
if (replident == REPLICA_IDENTITY_DEFAULT && OidIsValid(pkeyIndex) &&
!pkdeferrable)
relation->rd_replidindex = pkeyIndex;
and if there's no other suitable index, the result is that there is no
identity index for the table. So the change attached here should be consistent
with this approach.
--
Antonin Houska
Web: https://www.cybertec-postgresql.com
diff --git a/src/backend/commands/repack.c b/src/backend/commands/repack.c
index 67364cc60e3..8cfc3fde5c7 100644
--- a/src/backend/commands/repack.c
+++ b/src/backend/commands/repack.c
@@ -925,7 +925,8 @@ check_concurrent_repack_requirements(Relation rel, Oid *ident_idx_p)
* work and is not implemented yet.
*/
ident_idx = RelationGetReplicaIndex(rel);
- if (!OidIsValid(ident_idx) && OidIsValid(rel->rd_pkindex))
+ if (!OidIsValid(ident_idx) && OidIsValid(rel->rd_pkindex) &&
+ !rel->rd_ispkdeferrable)
ident_idx = rel->rd_pkindex;
if (!OidIsValid(ident_idx))
ereport(ERROR,
