statistics: do not load unnecessary index statistics

[Rustin170506]

What problem does this PR solve?

Issue Number: close #54022

Problem Summary:

See more at the issue.

A quick summary:
If you have tidb_opt_objective='determine' enabled, it will trigger some async stats load for some tables. If there tables histogram isn't in the system, it would cause some problems when we try to load it.

What changed and how does it work?

Check that the index statistics are really needed before loading them.

If we don't check it here, it will cause some problems if the index statistics don't exist.

Note: We still don't know why the histogram record for the index is missing. There are some possible issues:

1. When users create this table, they may not have created this index yet, so we don't create the record for it automatically. (But if this is the case, I don't why do we have the index in the stats cache information)
2. It may have been accidentally dropped, but I can't find any scenario where we would only drop one index for a table.

What happened in the test case:

1. We create a table and the statsHandle.Update(do.InfoSchema()) will load this table into the stats cache. You can test it by breakpoint here:
   
2. We drop the stats of the stats, it will clean up all system table records for this table.
3. We insert some data and wait for the modfiy_count and the count is not null in the mysql.stats_meta.
4. Try to select some data from this table by ID, it would trigger an async load. Then we get that bug.
   https://github.com/hi-rustin/tidb/blob/648fc6386aa9aabb365baf0912f1e18c4ab74d95/pkg/planner/core/logical_plan_builder.go#L4824
5. So in the test case we check if the index is still in the needed item array after we fix it.

Check List

Tests

• Unit test
• Integration test
• Manual test
• No need to test

  • I checked and no code files have been changed.

Side effects

• Performance regression: Consumes more CPU
• Performance regression: Consumes more Memory
• Breaking backward compatibility

Documentation

• Affects user behaviors
• Contains syntax changes
• Contains variable changes
• Contains experimental features
• Changes MySQL compatibility

Release note

Please refer to Release Notes Language Style Guide to write a quality release note.

Fix a memory leak when loading index statistics.
修复加载 index 统计信息可能会造成内存泄漏的问题

[codecov]

Codecov Report

Attention: Patch coverage is 18.75000% with 13 lines in your changes missing coverage. Please review.

Please upload report for BASE (release-7.5@1563cc5). Learn more about missing BASE report.

Additional details and impacted files

@@               Coverage Diff                @@
##             release-7.5     #54060   +/-   ##
================================================
  Coverage               ?   72.0278%           
================================================
  Files                  ?       1411           
  Lines                  ?     410758           
  Branches               ?          0           
================================================
  Hits                   ?     295860           
  Misses                 ?      94993           
  Partials               ?      19905           

Flag	Coverage Δ
unit	72.0278% <18.7500%> (?)

Flags with carried forward coverage won't be shown. Click here to find out more.

Components	Coverage Δ
dumpling	53.9913% <0.0000%> (?)
parser	∅ <0.0000%> (?)
br	53.4535% <0.0000%> (?)

[Rustin170506]

Tested locally:
After:

1. Start the cluster with patch: tiup playground v7.5.1 --db.binpath /Volumes/t7/code/tidb/bin/tidb-server
2. Create a table: create table if not exists t(a int, b int, index ia(a));
3. Drop the stats: drop stats t;
4. Insert some data: insert into t value(1,1), (2,2);
5. Wait for a while until we can see the stats: show stats_meta

mysql> show stats_meta
    -> ;
+---------+------------+----------------+---------------------+--------------+-----------+
| Db_name | Table_name | Partition_name | Update_time         | Modify_count | Row_count |
+---------+------------+----------------+---------------------+--------------+-----------+
| test    | t          |                | 2024-06-18 12:07:12 |            0 |         0 |
+---------+------------+----------------+---------------------+--------------+-----------+
1 row in set (0.01 sec)

6. Enable determinate: set tidb_opt_objective='determinate';
7. Try select data: select * from t where a = 1 and b = 1;

mysql> select * from t where a = 1 and b = 1;
+------+------+
| a    | b    |
+------+------+
|    1 |    1 |
+------+------+
1 row in set (0.01 sec)

8. Check logs:

[2024/06/18 12:09:47.346 +08:00] [WARN] [read.go:583] ["Although the index stats is not required to load, an attempt is still made to load it."] [table_id=102] [hist_id=1]

Before:

1. Start the cluster with patch: tiup playground v7.5.1 --db.binpath /Volumes/t7/code/tidb/bin/tidb-server
2. Create a table: create table if not exists t(a int, b int, index ia(a));
3. Drop the stats: drop stats t;
4. Insert some data: insert into t value(1,1), (2,2);
5. Wait for a while until we can see the stats: show stats_meta

mysql> show stats_meta
    -> ;
+---------+------------+----------------+---------------------+--------------+-----------+
| Db_name | Table_name | Partition_name | Update_time         | Modify_count | Row_count |
+---------+------------+----------------+---------------------+--------------+-----------+
| test    | t          |                | 2024-06-18 12:07:12 |            0 |         0 |
+---------+------------+----------------+---------------------+--------------+-----------+
1 row in set (0.01 sec)

6. Enable determinate: set tidb_opt_objective='determinate';
7. Try select data: select * from t where a = 1 and b = 1;

mysql> select * from t where a = 1 and b = 1;
+------+------+
| a    | b    |
+------+------+
|    1 |    1 |
+------+------+
1 row in set (0.01 sec)

8. Check logs:

[2024/06/18 12:15:01.729 +08:00] [ERROR] [read.go:603] ["fail to get stats version for this histogram"] [table_id=102] [hist_id=1]
[2024/06/18 12:15:04.728 +08:00] [ERROR] [read.go:603] ["fail to get stats version for this histogram"] [table_id=102] [hist_id=1]

[Rustin170506]

🔢 Self-check (PR reviewed by myself and ready for feedback.)

[Rustin170506]

🔢 Self-check (PR reviewed by myself and ready for feedback.)

[elsa0520]

LGTM

[hawkingrei]

Tested locally: After:

1. Start the cluster with patch: `tiup playground v7.5.1  --db.binpath /Volumes/t7/code/tidb/bin/tidb-server`

2. Create a table: `create table if not exists t(a int, b int, index ia(a));`

3. Drop the stats: `drop stats t;`

4. Insert some data: `insert into t value(1,1), (2,2);`

5. Wait for a while until we can see the stats: `show stats_meta`

mysql> show stats_meta
    -> ;
+---------+------------+----------------+---------------------+--------------+-----------+
| Db_name | Table_name | Partition_name | Update_time         | Modify_count | Row_count |
+---------+------------+----------------+---------------------+--------------+-----------+
| test    | t          |                | 2024-06-18 12:07:12 |            0 |         0 |
+---------+------------+----------------+---------------------+--------------+-----------+
1 row in set (0.01 sec)

6. Enable determinate: `set tidb_opt_objective='determinate';`

7. Try select data: `select * from t where a = 1 and b = 1;`

mysql> select * from t where a = 1 and b = 1;
+------+------+
| a    | b    |
+------+------+
|    1 |    1 |
+------+------+
1 row in set (0.01 sec)

8. Check logs:

[2024/06/18 12:09:47.346 +08:00] [WARN] [read.go:583] ["Although the index stats is not required to load, an attempt is still made to load it."] [table_id=102] [hist_id=1]

Before:

1. Start the cluster with patch: `tiup playground v7.5.1  --db.binpath /Volumes/t7/code/tidb/bin/tidb-server`

2. Create a table: `create table if not exists t(a int, b int, index ia(a));`

3. Drop the stats: `drop stats t;`

4. Insert some data: `insert into t value(1,1), (2,2);`

5. Wait for a while until we can see the stats: `show stats_meta`

mysql> show stats_meta
    -> ;
+---------+------------+----------------+---------------------+--------------+-----------+
| Db_name | Table_name | Partition_name | Update_time         | Modify_count | Row_count |
+---------+------------+----------------+---------------------+--------------+-----------+
| test    | t          |                | 2024-06-18 12:07:12 |            0 |         0 |
+---------+------------+----------------+---------------------+--------------+-----------+
1 row in set (0.01 sec)

6. Enable determinate: `set tidb_opt_objective='determinate';`

7. Try select data: `select * from t where a = 1 and b = 1;`

mysql> select * from t where a = 1 and b = 1;
+------+------+
| a    | b    |
+------+------+
|    1 |    1 |
+------+------+
1 row in set (0.01 sec)

8. Check logs:

[2024/06/18 12:15:01.729 +08:00] [ERROR] [read.go:603] ["fail to get stats version for this histogram"] [table_id=102] [hist_id=1]
[2024/06/18 12:15:04.728 +08:00] [ERROR] [read.go:603] ["fail to get stats version for this histogram"] [table_id=102] [hist_id=1]

You can add this desciption into Manual test.

[time-and-fate]

We can use this to trigger it immediately.

Suggested change

	require.Eventually(t, func() bool {
	rows := tk.MustQuery("show stats_meta").Rows()
	return len(rows) > 0
	}, 2*time.Minute, 5*time.Millisecond)
	h := dom.StatsHandle()
	require.NoError(t, h.DumpStatsDeltaToKV(true))
	require.NoError(t, h.Update(dom.InfoSchema()))

[Rustin170506]

Nice! I forgot we can directly call an update here.

[Rustin170506]

Updated.

[ti-chi-bot]

[APPROVALNOTIFIER] This PR is APPROVED

This pull-request has been approved by: elsa0520, time-and-fate

The full list of commands accepted by this bot can be found here.

The pull request process is described here

Needs approval from an approver in each of these files:

• OWNERS [time-and-fate]

Approvers can indicate their approval by writing /approve in a comment
Approvers can cancel approval by writing /approve cancel in a comment

[ti-chi-bot]

[LGTM Timeline notifier]

Timeline:

• 2024-06-18 06:54:50.954675945 +0000 UTC m=+97817.440164778: ☑️ agreed by elsa0520.
• 2024-06-18 08:23:05.483640997 +0000 UTC m=+103111.969129828: ☑️ agreed by time-and-fate.

[ti-chi-bot]

In response to a cherrypick label: new pull request created to branch release-7.1: #54087.