Details
-
Bug
-
Status: Closed (View Workflow)
-
Major
-
Resolution: Fixed
-
10.2, 10.3, 10.4, 10.5, 10.6
-
None
Description
When online alter rollbacks due to MDL time out, it doesn't mark the index online status as ONLINE_INDEX_ABORTED. Concurrent update fails to update the secondary index because the index is being dropped.
The following test case demonstrates the problem
diff --git a/storage/innobase/row/row0upd.cc b/storage/innobase/row/row0upd.cc
|
index 10156c7d3ab..39c110c2df2 100644
|
--- a/storage/innobase/row/row0upd.cc
|
+++ b/storage/innobase/row/row0upd.cc
|
@@ -2517,6 +2517,8 @@ row_upd_sec_index_entry(
|
|
mem_heap_empty(heap);
|
|
+ DEBUG_SYNC_C_IF_THD(trx->mysql_thd,
|
+ "before_row_upd_sec_new_index_entry");
|
/* Build a new index entry */
|
entry = row_build_index_entry(node->upd_row, node->upd_ext,
|
index, heap);
|
Test case:
--source include/have_innodb.inc
|
create table t1(f1 char(10), f2 char(10) not null, f3 int not null,
|
f4 int not null, primary key(f3))engine=innodb;
|
insert into t1 values('a','a', 1, 1), ('b','b', 2, 2), ('c', 'c', 3, 3), ('d', 'd', 4, 4);
|
SET DEBUG_SYNC="row_merge_after_scan SIGNAL con1_start WAIT_FOR con1_insert";
|
SET DEBUG_SYNC="innodb_commit_inplace_alter_table_wait SIGNAL con1_wait WAIT_FOR con1_update";
|
send ALTER TABLE t1 ADD UNIQUE INDEX(f1(3), f4), ADD UNIQUE INDEX(f2), ALGORITHM=INPLACE;
|
|
|
connect(con1,localhost,root,,,);
|
SET DEBUG_SYNC="now WAIT_FOR con1_start";
|
begin;
|
INSERT INTO t1 VALUES('e','e',5, 5);
|
SET DEBUG_SYNC="now SIGNAL con1_insert";
|
SET DEBUG_SYNC="now WAIT_FOR con1_wait";
|
SET DEBUG_SYNC="before_row_upd_sec_new_index_entry SIGNAL con1_update WAIT_FOR alter_rollback";
|
SEND UPDATE t1 set f4 = 10 order by f1 desc limit 2;
|
|
|
connection default;
|
--error 1205
|
reap;
|
SET DEBUG_SYNC="now SIGNAL alter_rollback";
|
|
|
connection con1;
|
reap;
|
commit;
|
|
|
connection default;
|
disconnect con1;
|
DROP TABLE t1;
|
SET DEBUG_SYNC="RESET";
|
.opt file:
--lock_wait_timeout=2
|
The assert failure is
Thread 1 (Thread 0x7fce7c4a4700 (LWP 29920)):
|
#0 __pthread_kill (threadid=<optimized out>, signo=6) at ../sysdeps/unix/sysv/linux/pthread_kill.c:57
|
#1 0x0000563735d38299 in my_write_core (sig=6) at /home/thiru/mariarepo/10.2/10.2-test/mysys/stacktrace.c:382
|
#2 0x00005637355b12fc in handle_fatal_signal (sig=6) at /home/thiru/mariarepo/10.2/10.2-test/sql/signal_handler.cc:343
|
#3 <signal handler called>
|
#4 __GI_raise (sig=sig@entry=6) at ../sysdeps/unix/sysv/linux/raise.c:51
|
#5 0x00007fce81eb28b1 in __GI_abort () at abort.c:79
|
#6 0x00007fce81ea242a in __assert_fail_base (fmt=0x7fce82029a38 "%s%s%s:%u: %s%sAssertion `%s' failed.\n%n", assertion=assertion@entry=0x563735f8b7ca "col->ord_part", file=file@entry=0x563735f8b620 "/home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0row.cc", line=line@entry=294, function=function@entry=0x563735f8bd20 <row_build_index_entry_low(dtuple_t const*, row_ext_t const*, dict_index_t*, mem_block_info_t*, unsigned long)::__PRETTY_FUNCTION__> "dtuple_t* row_build_index_entry_low(const dtuple_t*, const row_ext_t*, dict_index_t*, mem_heap_t*, ulint)") at assert.c:92
|
#7 0x00007fce81ea24a2 in __GI___assert_fail (assertion=0x563735f8b7ca "col->ord_part", file=0x563735f8b620 "/home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0row.cc", line=294, function=0x563735f8bd20 <row_build_index_entry_low(dtuple_t const*, row_ext_t const*, dict_index_t*, mem_block_info_t*, unsigned long)::__PRETTY_FUNCTION__> "dtuple_t* row_build_index_entry_low(const dtuple_t*, const row_ext_t*, dict_index_t*, mem_heap_t*, ulint)") at assert.c:101
|
#8 0x000056373590ef10 in row_build_index_entry_low (row=0x7fce2801d4d0, ext=0x0, index=0x7fce240a18e0, heap=0x7fce2801de60, flag=0) at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0row.cc:294
|
#9 0x000056373593a995 in row_build_index_entry (row=0x7fce2801d4d0, ext=0x0, index=0x7fce240a18e0, heap=0x7fce2801de60) at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/include/row0row.ic:129
|
#10 0x0000563735940316 in row_upd_sec_index_entry (node=0x7fce2801c280, thr=0x7fce2801c5e8) at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0upd.cc:2523
|
#11 0x0000563735940524 in row_upd_sec_step (node=0x7fce2801c280, thr=0x7fce2801c5e8) at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0upd.cc:2555
|
#12 0x00005637359427a9 in row_upd (node=0x7fce2801c280, thr=0x7fce2801c5e8) at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0upd.cc:3317
|
#13 0x0000563735942afd in row_upd_step (thr=0x7fce2801c5e8) at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0upd.cc:3434
|
#14 0x00005637358eccf5 in row_update_for_mysql (prebuilt=0x7fce2801b710) at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/row/row0mysql.cc:1825
|
#15 0x00005637357b2b07 in ha_innobase::update_row (this=0x7fce2801ab28, old_row=0x7fce2801a6f0 "\375e e \005", new_row=0x7fce2801a6d0 "\375e e \005") at /home/thiru/mariarepo/10.2/10.2-test/storage/innobase/handler/ha_innodb.cc:8960
|
#16 0x00005637355c0cec in handler::ha_update_row (this=0x7fce2801ab28, old_data=0x7fce2801a6f0 "\375e e \005", new_data=0x7fce2801a6d0 "\375e e \005") at /home/thiru/mariarepo/10.2/10.2-test/sql/handler.cc:6148
|
#17 0x000056373542b2ff in mysql_update (thd=0x7fce28000d50, table_list=0x7fce280113f8, fields=..., values=..., conds=0x0, order_num=1, order=0x7fce28011cf0, limit=2, handle_duplicates=DUP_ERROR, ignore=false, found_return=0x7fce7c4a2c50, updated_return=0x7fce7c4a2d00) at /home/thiru/mariarepo/10.2/10.2-test/sql/sql_update.cc:819
|
#18 0x000056373533b530 in mysql_execute_command (thd=0x7fce28000d50) at /home/thiru/mariarepo/10.2/10.2-test/sql/sql_parse.cc:4032
|
#19 0x0000563735347215 in mysql_parse (thd=0x7fce28000d50, rawbuf=0x7fce280112f8 "UPDATE t1 set f4 = 10 order by f1 desc limit 2", length=46, parser_state=0x7fce7c4a35f0, is_com_multi=false, is_next_command=false) at /home/thiru/mariarepo/10.2/10.2-test/sql/sql_parse.cc:7762
|
#20 0x0000563735335619 in dispatch_command (command=COM_QUERY, thd=0x7fce28000d50, packet=0x7fce28008b11 "UPDATE t1 set f4 = 10 order by f1 desc limit 2", packet_length=46, is_com_multi=false, is_next_command=false) at /home/thiru/mariarepo/10.2/10.2-test/sql/sql_parse.cc:1827
|
#21 0x000056373533409e in do_command (thd=0x7fce28000d50) at /home/thiru/mariarepo/10.2/10.2-test/sql/sql_parse.cc:1381
|
#22 0x00005637354897db in do_handle_one_connection (connect=0x563737e0ad30) at /home/thiru/mariarepo/10.2/10.2-test/sql/sql_connect.cc:1336
|
#23 0x0000563735489546 in handle_one_connection (arg=0x563737e0ad30) at /home/thiru/mariarepo/10.2/10.2-test/sql/sql_connect.cc:1241
|
#24 0x0000563735c8b976 in pfs_spawn_thread (arg=0x563737e4b420) at /home/thiru/mariarepo/10.2/10.2-test/storage/perfschema/pfs.cc:1869
|
#25 0x00007fce82bf46db in start_thread (arg=0x7fce7c4a4700) at pthread_create.c:463
|
#26 0x00007fce81f93a3f in clone () at ../sysdeps/unix/sysv/linux/x86_64/clone.S:95
|
I think InnoDB should fix this by taking InnoDB table lock during rollback of alter table. In that case, we can avoid concurrent DML. Maybe having some extra() to indicate MDL time out and take X-lock?
Open to suggestion for the solution though.