the time returned to the client is inconsistent with the ddl job info for truncate table

[aytrack]

Bug Report

Please answer these questions before submitting your issue. Thanks!

1. Minimal reproduce step (Required)

1. tiup playground nightly --tiflash 0

create table t02(a int, b int, primary key (a)) partition by hash(a) partitions 1000;
truncate table t02;
truncate table t02;
truncate table t02;

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

the time for truncate table from ddl jobs info should consistent with client's time

3. What did you see instead (Required)

you can see that client wait a long time but the ddl info record only a few time.

4. What is your TiDB version? (Required)

[16:34:51]TiDB root:test> select tidb_version();
+-----------------------------------------------------------+
| tidb_version()                                            |
+-----------------------------------------------------------+
| Release Version: v7.2.0-alpha                             |
| Edition: Community                                        |
| Git Commit Hash: 8633c8ce9921bf17a92b582a6b5d299d95213ac6 |
| Git Branch: heads/refs/tags/v7.2.0-alpha                  |
| UTC Build Time: 2023-06-15 14:25:30                       |
| GoVersion: go1.20.3                                       |
| Race Enabled: false                                       |
| Check Table Before Drop: false                            |
| Store: tikv                                               |
+-----------------------------------------------------------+

[mjonss]

I would propose to move the preSplitAndScatter call into the ddl worker instead, probably directly after the new table/partitions are created, before they are public to be written.

[crazycs520]

I don't think this requirement is reasonable, and we don't promise that the DDL job time is equal to the execution time of the DDL SQL statement. DDL job execution time is only a part, not the whole time, of DDL SQL statement execution time.