Noisy messages for index histogram

Bug Report

Noisy messages like this:

{
  "level": "WARN",
  "time": "2025/10/08 18:10:37.342 +00:00",
  "caller": "stats_syncload.go:415",
  "message": "fail to get hist meta for this histogram, possibly a deleted one",
  "table_id": 1039,
  "hist_id": 2,
  "is_index": true
}

This happens for multiple table_id's and multiple hist_id's, but is_index is always true.

It looks like mysql.stats_histograms doesn't have an entry for the table_id/hist_id/is_index combination.

The message seems to repeat with the same properties.

This seems similar to:

1. Minimal reproduce step (Required)

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

3. What did you see instead (Required)

4. What is your TiDB version? (Required)

v8.5.2

The reproduction step:

Turn off the auto analyze:

set global tidb_enable_auto_analyze = false;

Create a table and insert some data:

set @@cte_max_recursion_depth=10000;
use test;
create table h (idx int, code int, typ1 int, typ2 int, update_time int, key k1(idx, typ1, typ2), key k2(idx, update_time));
insert into h select * from (
  with recursive tt as (
    select 0 idx, 0 as code, 0 as typ1, 0 as typ2, 0 as update_time
    union all
    select mod(update_time, 100) as idx, 0 as code, 0 as typ1, 0 as typ2, update_time+1 as update_time from tt where update_time<5000
  ) select * from tt
) tt;

Analyze the table and explain to check the stats status:

analyze table h;
explain select * from h where idx = 0;

Add a new index:

ALTER TABLE h ADD INDEX k3(code);
explain select * from h where code = 0;

Check logs:

[2025/10/14 12:49:01.140 +02:00] [WARN] [stats_syncload.go:416] ["fail to get hist meta for this histogram, possibly a deleted one"] [table_id=110] [hist_id=3] [is_index=true]

I think we should make the error message clearer and use a sample logger to reduce the amount of logs.

~~The load behavior is expected, but it’s just too overwhelming.~~

There are some subtle differences between the master branch and 8.5.2. Further investigation is needed.

On 8.5.2, I can confirm that the issue is from here:

The checked field is always false.

The reason the checked field is always false is that the lite init stats process never sets this field. Additionally, during the stats update process, there are no updates after the last histogram version, so the columns’ and indexes’ statistics are never checked. As a result, this field remains false.

What about v8.5.3, does that have the same issue?

I have confirmed that this issue still exists even in the latest release-8.5 branch.

For the master branch, I believe this issue still existed before we merged PR #63626.

statistics: use sampled logger and improve error messages (pingcap#63962

planner: pick #64622 #63519 #63962 #64644 #64501 (#64689)

statistics: use sampled logger and improve error messages (#63962) (#…

Navigation Menu

Repository navigation

Noisy messages for index histogram #63880

Bug Report

1. Minimal reproduce step (Required)

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

3. What did you see instead (Required)

4. What is your TiDB version? (Required)

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Noisy messages for index histogram #63880

Description

Bug Report

1. Minimal reproduce step (Required)

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

3. What did you see instead (Required)

4. What is your TiDB version? (Required)

Activity

0xPoe commented on Oct 14, 2025

0xPoe commented on Oct 14, 2025

0xPoe commented on Oct 14, 2025

0xPoe commented on Oct 16, 2025

dveeden commented on Oct 16, 2025

0xPoe commented on Oct 16, 2025

0xPoe commented on Oct 16, 2025

Metadata

Metadata

Assignees

Labels

Type

Projects

Milestone

Relationships

Development

Issue actions