From e01023828046a6b517109c255a07765f9bc0a3fb Mon Sep 17 00:00:00 2001 From: Dimitri Fontaine Date: Tue, 9 Jan 2018 21:04:11 +0100 Subject: [PATCH] Implement ALTER TABLE ... RENAME TO ... MIME-Version: 1.0 Content-Type: text/plain; charset=UTF-8 Content-Transfer-Encoding: 8bit The implementation was already mostly in place, but the code was protected by a principled check against the operation. Turns out there's a nasty concurrency bug though with long identifier names, much as in #1664. To prevent deadlocks from happening, we could either review the DDL transaction management in shards and placements, or we can simply reject names with (NAMEDATALEN - 1) chars or more — that's because of the PostgreSQL array types being created with a one-char prefix: '_'. --- .../distributed/executor/multi_utility.c | 7 +- .../distributed/relay/relay_event_utility.c | 25 +++++++ .../regress/expected/multi_name_lengths.out | 3 +- .../expected/multi_reference_table.out | 7 +- .../input/multi_alter_table_statements.source | 25 ++++++- .../multi_alter_table_statements.source | 68 +++++++++++++++++-- .../regress/sql/multi_reference_table.sql | 7 +- 7 files changed, 123 insertions(+), 19 deletions(-) diff --git a/src/backend/distributed/executor/multi_utility.c b/src/backend/distributed/executor/multi_utility.c index d39ad7ef1..519954eea 100644 --- a/src/backend/distributed/executor/multi_utility.c +++ b/src/backend/distributed/executor/multi_utility.c @@ -2740,12 +2740,7 @@ ErrorIfUnsupportedRenameStmt(RenameStmt *renameStmt) { Assert(IsAlterTableRenameStmt(renameStmt)); - if (renameStmt->renameType == OBJECT_TABLE) - { - ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), - errmsg("renaming distributed tables is currently unsupported"))); - } - else if (renameStmt->renameType == OBJECT_TABCONSTRAINT) + if (renameStmt->renameType == OBJECT_TABCONSTRAINT) { ereport(ERROR, (errcode(ERRCODE_FEATURE_NOT_SUPPORTED), errmsg("renaming constraints belonging to distributed tables is " diff --git a/src/backend/distributed/relay/relay_event_utility.c b/src/backend/distributed/relay/relay_event_utility.c index 827e737af..120e74999 100644 --- a/src/backend/distributed/relay/relay_event_utility.c +++ b/src/backend/distributed/relay/relay_event_utility.c @@ -388,12 +388,37 @@ RelayEventExtendNames(Node *parseTree, char *schemaName, uint64 shardId) char **oldRelationName = &(renameStmt->relation->relname); char **newRelationName = &(renameStmt->newname); char **objectSchemaName = &(renameStmt->relation->schemaname); + int newRelationNameLength; /* prefix with schema name if it is not added already */ SetSchemaNameIfNotExist(objectSchemaName, schemaName); AppendShardIdToName(oldRelationName, shardId); AppendShardIdToName(newRelationName, shardId); + + /* + * PostgreSQL creates array types for each ordinary table, with + * the same name plus a prefix of '_'. + * + * ALTER TABLE ... RENAME TO ... also renames the underlying + * array type, and the DDL is run in parallel connections over + * all the placements and shards at once. Concurrent access + * here deadlocks. + * + * Let's provide an easier to understand error message here + * than the deadlock one. + * + * See also https://github.com/citusdata/citus/issues/1664 + */ + newRelationNameLength = strlen(*newRelationName); + if (newRelationNameLength >= (NAMEDATALEN - 1)) + { + ereport(ERROR, + (errcode(ERRCODE_NAME_TOO_LONG), + errmsg( + "shard name %s exceeds %d characters", + *newRelationName, NAMEDATALEN - 1))); + } } else if (objectType == OBJECT_COLUMN || objectType == OBJECT_TRIGGER) { diff --git a/src/test/regress/expected/multi_name_lengths.out b/src/test/regress/expected/multi_name_lengths.out index e57cd137a..7cf9d11d7 100644 --- a/src/test/regress/expected/multi_name_lengths.out +++ b/src/test/regress/expected/multi_name_lengths.out @@ -113,7 +113,8 @@ SELECT "Constraint", "Definition" FROM table_checks WHERE relid='public.name_len \c - - - :master_port -- Placeholders for RENAME operations ALTER TABLE name_lengths RENAME TO name_len_12345678901234567890123456789012345678901234567890; -ERROR: renaming distributed tables is currently unsupported +ERROR: shard name name_len_12345678901234567890123456789012345678_fcd8ab6f_225002 exceeds 63 characters +CONTEXT: while executing command on localhost:57638 ALTER TABLE name_lengths RENAME CONSTRAINT unique_12345678901234567890123456789012345678901234567890 TO unique2_12345678901234567890123456789012345678901234567890; ERROR: renaming constraints belonging to distributed tables is currently unsupported -- Verify that CREATE INDEX on already distributed table has proper shard names. diff --git a/src/test/regress/expected/multi_reference_table.out b/src/test/regress/expected/multi_reference_table.out index 76df77e19..073104d8f 100644 --- a/src/test/regress/expected/multi_reference_table.out +++ b/src/test/regress/expected/multi_reference_table.out @@ -1382,12 +1382,13 @@ SELECT "Column", "Type", "Modifiers" FROM table_desc WHERE relid='public.referen (0 rows) \c - - - :master_port --- as we expect, renaming and setting WITH OIDS does not work for reference tables -ALTER TABLE reference_table_ddl RENAME TO reference_table_ddl_test; -ERROR: renaming distributed tables is currently unsupported +-- as we expect, setting WITH OIDS does not work for reference tables ALTER TABLE reference_table_ddl SET WITH OIDS; ERROR: alter table command is currently unsupported DETAIL: Only ADD|DROP COLUMN, SET|DROP NOT NULL, SET|DROP DEFAULT, ADD|DROP CONSTRAINT, ATTACH|DETACH PARTITION and TYPE subcommands are supported. +-- now test the renaming of the table, and back to the expected name +ALTER TABLE reference_table_ddl RENAME TO reference_table_ddl_test; +ALTER TABLE reference_table_ddl_test RENAME TO reference_table_ddl; -- now test reference tables against some helper UDFs that Citus provides -- cannot delete / drop shards from a reference table SELECT master_apply_delete_command('DELETE FROM reference_table_ddl'); diff --git a/src/test/regress/input/multi_alter_table_statements.source b/src/test/regress/input/multi_alter_table_statements.source index 099c5db88..2cefeaf88 100644 --- a/src/test/regress/input/multi_alter_table_statements.source +++ b/src/test/regress/input/multi_alter_table_statements.source @@ -148,9 +148,8 @@ ALTER TABLE lineitem_alter ADD COLUMN new_column non_existent_type; ALTER TABLE lineitem_alter ALTER COLUMN null_column SET NOT NULL; ALTER TABLE lineitem_alter ALTER COLUMN l_partkey SET DEFAULT 'a'; --- Verify that we error out on non-column RENAME statements +-- Verify that we error out on RENAME CONSTRAINT statement -ALTER TABLE lineitem_alter RENAME TO lineitem_renamed; ALTER TABLE lineitem_alter RENAME CONSTRAINT constraint_a TO constraint_b; -- Verify that IF EXISTS works as expected with RENAME statements @@ -356,10 +355,30 @@ FROM ORDER BY attnum; \c - - - :master_port +-- verify that we can rename distributed tables +SHOW citus.enable_ddl_propagation; +ALTER TABLE lineitem_alter RENAME TO lineitem_renamed; + +-- verify rename is performed +SELECT relname FROM pg_class WHERE relname = 'lineitem_renamed'; + +-- show rename worked on one worker, too +\c - - - :worker_1_port +SELECT relname FROM pg_class WHERE relname LIKE 'lineitem_renamed%' ORDER BY relname; +\c - - - :master_port + +-- revert it to original name +ALTER TABLE lineitem_renamed RENAME TO lineitem_alter; + +-- show rename worked on one worker, too +\c - - - :worker_1_port +SELECT relname FROM pg_class WHERE relname LIKE 'lineitem_alter%' ORDER BY relname; +\c - - - :master_port + -- verify that we don't intercept DDL commands if propagation is turned off SET citus.enable_ddl_propagation to false; --- table rename statement can be performed now +-- table rename statement can be performed on the coordinator only now ALTER TABLE lineitem_alter RENAME TO lineitem_renamed; -- verify rename is performed SELECT relname FROM pg_class WHERE relname = 'lineitem_alter' or relname = 'lineitem_renamed'; diff --git a/src/test/regress/output/multi_alter_table_statements.source b/src/test/regress/output/multi_alter_table_statements.source index 1dd7973b9..49017ff2e 100644 --- a/src/test/regress/output/multi_alter_table_statements.source +++ b/src/test/regress/output/multi_alter_table_statements.source @@ -377,9 +377,7 @@ ERROR: column "null_column" contains null values CONTEXT: while executing command on localhost:57638 ALTER TABLE lineitem_alter ALTER COLUMN l_partkey SET DEFAULT 'a'; ERROR: invalid input syntax for integer: "a" --- Verify that we error out on non-column RENAME statements -ALTER TABLE lineitem_alter RENAME TO lineitem_renamed; -ERROR: renaming distributed tables is currently unsupported +-- Verify that we error out on RENAME CONSTRAINT statement ALTER TABLE lineitem_alter RENAME CONSTRAINT constraint_a TO constraint_b; ERROR: renaming constraints belonging to distributed tables is currently unsupported -- Verify that IF EXISTS works as expected with RENAME statements @@ -783,10 +781,72 @@ ORDER BY attnum; ........pg.dropped.24........ | - (30 rows) +\c - - - :master_port +-- verify that we can rename distributed tables +SHOW citus.enable_ddl_propagation; + citus.enable_ddl_propagation +------------------------------ + on +(1 row) + +ALTER TABLE lineitem_alter RENAME TO lineitem_renamed; +-- verify rename is performed +SELECT relname FROM pg_class WHERE relname = 'lineitem_renamed'; + relname +------------------ + lineitem_renamed +(1 row) + +-- show rename worked on one worker, too +\c - - - :worker_1_port +SELECT relname FROM pg_class WHERE relname LIKE 'lineitem_renamed%' ORDER BY relname; + relname +------------------------- + lineitem_renamed_220000 + lineitem_renamed_220001 + lineitem_renamed_220002 + lineitem_renamed_220003 + lineitem_renamed_220004 + lineitem_renamed_220005 + lineitem_renamed_220006 + lineitem_renamed_220007 + lineitem_renamed_220008 + lineitem_renamed_220010 + lineitem_renamed_220011 + lineitem_renamed_220012 + lineitem_renamed_220013 + lineitem_renamed_220014 +(14 rows) + +\c - - - :master_port +-- revert it to original name +ALTER TABLE lineitem_renamed RENAME TO lineitem_alter; +-- show rename worked on one worker, too +\c - - - :worker_1_port +SELECT relname FROM pg_class WHERE relname LIKE 'lineitem_alter%' ORDER BY relname; + relname +----------------------- + lineitem_alter_220000 + lineitem_alter_220001 + lineitem_alter_220002 + lineitem_alter_220003 + lineitem_alter_220004 + lineitem_alter_220005 + lineitem_alter_220006 + lineitem_alter_220007 + lineitem_alter_220008 + lineitem_alter_220009 + lineitem_alter_220010 + lineitem_alter_220011 + lineitem_alter_220012 + lineitem_alter_220013 + lineitem_alter_220014 +(15 rows) + \c - - - :master_port -- verify that we don't intercept DDL commands if propagation is turned off SET citus.enable_ddl_propagation to false; --- table rename statement can be performed now +-- table rename statement can be performed on the coordinator only now ALTER TABLE lineitem_alter RENAME TO lineitem_renamed; -- verify rename is performed SELECT relname FROM pg_class WHERE relname = 'lineitem_alter' or relname = 'lineitem_renamed'; diff --git a/src/test/regress/sql/multi_reference_table.sql b/src/test/regress/sql/multi_reference_table.sql index ede3a7956..79881ad22 100644 --- a/src/test/regress/sql/multi_reference_table.sql +++ b/src/test/regress/sql/multi_reference_table.sql @@ -872,10 +872,13 @@ SELECT "Column", "Type", "Modifiers" FROM table_desc WHERE relid='public.referen \di reference_index_2* \c - - - :master_port --- as we expect, renaming and setting WITH OIDS does not work for reference tables -ALTER TABLE reference_table_ddl RENAME TO reference_table_ddl_test; +-- as we expect, setting WITH OIDS does not work for reference tables ALTER TABLE reference_table_ddl SET WITH OIDS; +-- now test the renaming of the table, and back to the expected name +ALTER TABLE reference_table_ddl RENAME TO reference_table_ddl_test; +ALTER TABLE reference_table_ddl_test RENAME TO reference_table_ddl; + -- now test reference tables against some helper UDFs that Citus provides -- cannot delete / drop shards from a reference table