tikv aborts txn: KeyTooLarge

[wjhuang2016]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

Enable the new collation.

drop table if exists t;
create table t(a varchar(700) collate utf8mb4_unicode_ci, key (a));
insert into t values (repeat("ﷻ", 700));

mysql> show create table t;
+-------+---------------------------------------------------------------------------------------------------------------------------------+
| Table | Create Table                                                                                                                    |
+-------+---------------------------------------------------------------------------------------------------------------------------------+
| t     | CREATE TABLE `t` (
  `a` varchar(700) DEFAULT NULL,
  KEY `a` (`a`)
) ENGINE=InnoDB DEFAULT CHARSET=utf8mb4 COLLATE=utf8mb4_bin |
+-------+---------------------------------------------------------------------------------------------------------------------------------+
1 row in set (0.00 sec)

mysql> drop table t;
Query OK, 0 rows affected (0.25 sec)

mysql> create table t(a varchar(700) collate utf8mb4_unicode_ci, key (a));
Query OK, 0 rows affected (0.11 sec)

mysql> insert into t values (repeat("ﷻ", 700));
ERROR 1105 (HY000): tikv aborts txn: KeyTooLarge(14220, 4096)

2. What did you expect to see? (Required)

No error.

3. What did you see instead (Required)

ERROR 1105 (HY000): tikv aborts txn: KeyTooLarge(14220, 4096)

4. What is your TiDB version? (Required)

master

[tangenta]

We also need to check the clustered primary key length:

create table t(a varchar(700) collate utf8mb4_unicode_ci, primary key (a) clustered);
insert into t values (repeat("ﷻ", 700));

ERROR 1105 (HY000): tikv aborts txn: KeyTooLarge(14202, 4096)

The prefix of key(table ID, index ID...) and sort key length should also be considered too.

[wjhuang2016]

Since it hits the TiKV's limit, it's expected to return an error.

[github-actions]

Please check whether the issue should be labeled with 'affects-x.y' or 'backport-x.y.z',
and then remove 'needs-more-info' label.