MySQL query optimizer doesn't necessarily use unique key in BuildUniqueKeyRangeEndPreparedQueryViaOffset's query

Ouverte
  • Quelque chose ne va pas ?
#1,538 1 commentaire 2 réactions 0 personnes assignées Voir sur GitHub

Les mainteneurs répondent en général sous 10 jours

Personne n'a encore pris cette issue.

Évaluation

Difficulté
4/5
Temps estimé
3-5 jours
Accessibilité débutants
35/100
Type d'issue
Bug
Clarté
Plutôt claire
Activité
À l'abandon
Stack technique
go, mysql
Domaine
backend, databases

Piste de recherche

Recherchez BuildUniqueKeyRangeEndPreparedQueryViaOffset et examinez comment cette requête MySQL est construite. Reproduisez le comportement signalé de EXPLAIN et SELECT avec et sans USE INDEX (PRIMARY), puis déterminez s’il est sûr de forcer l’index. Le travail est terminé lorsque le problème de l’optimiseur est résolu avec une couverture pour le chemin de requête concerné, si cette approche est appropriée.

Rédigé par le modèle d'indexation à partir du texte de l'issue.

Description

bug performance

We've run into an issue where the gh-ost migration starts off fine, but at some point, the MySQL query optimizer behavior changes for this query, and this query starts taking ~15m.

The explain query shows that the MySQL query optimizer is doing an index scan

mysql> explain select `applicationId`, `id` from `_workflowActionExecutions` where ((`applicationId` > _binary'appAAAA') or (((`applicationId` = _binary'appAAAA')) AND (`id` > _binary'waxAAAA'))) and ((`applicationId` < _binary'appZZZZ') or (((`applicationId` =
_binary'appZZZZ')) AND (`id` < _binary'waxZZZZ')) or ((`applicationId` = _binary'appZZZZ') and (`id` = _binary'waxZZZZ'))) order by `applicationId` asc, `id` asc limit 1 offset 999;
+----+-------------+---------------------------+------------+-------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------+---------+------+------+----------+--------------------------+
| id | select_type | table                     | partitions | type  | possible_keys                                                                                                                                                                                       | key     | key_len | ref  | rows | filtered | Extra                    |
+----+-------------+---------------------------+------------+-------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------+---------+------+------+----------+--------------------------+
|  1 | SIMPLE      | _workflowActionExecutions | NULL       | index | PRIMARY,IDX_applicationId_workflowGraphExecutionId,IDX_applicationId_workflowExecutionId_status_branchId,IDX_applicationId_status_workflowId,IDX_applicationId_workflowId_status_processTime,IDX_id | PRIMARY | 34      | NULL | 1000 |   100.00 | Using where; Using index |
+----+-------------+---------------------------+------------+-------+-----------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------+---------+---------+------+------+----------+--------------------------+
1 row in set, 1 warning (0.47 sec)

(ids fuzzed)

This is because MySQL thinks that if we USE INDEX (PRIMARY), it will end up doing a massive range scan

mysql> explain select `applicationId`, `id`
    -> from `_workflowActionExecutions` USE INDEX (PRIMARY)
    -> where ((`applicationId` > _binary'appAAAA') or (((`applicationId` = _binary'appAAAA')) AND (`id` > _binary'waxAAAA'))) and ((`applicationId` < _binary'appZZZZ') or (((`applicationId` = _binary'appZZZZ')) AND (`id` < _binary'waxZZZZ')) or ((`applicationId` = _binary'appZZZZ') and (`id` = _binary'waxZZZZ')))
    -> order by `applicationId` asc, `id` asc
    -> limit 1
    -> offset 999;
+----+-------------+---------------------------+------------+-------+---------------+---------+---------+------+----------+----------+--------------------------+
| id | select_type | table                     | partitions | type  | possible_keys | key     | key_len | ref  | rows     | filtered | Extra                    |
+----+-------------+---------------------------+------------+-------+---------------+---------+---------+------+----------+----------+--------------------------+
|  1 | SIMPLE      | _workflowActionExecutions | NULL       | range | PRIMARY       | PRIMARY | 34      | NULL | 95809902 |   100.00 | Using where; Using index |
+----+-------------+---------------------------+------------+-------+---------------+---------+---------+------+----------+----------+--------------------------+
1 row in set, 1 warning (0.46 sec)

(note the massive amount of rows)

But if I run the query with USE INDEX (PRIMARY), it's actually really fast, instead of taking ~15m

mysql> select `applicationId`, `id`
    -> from `_workflowActionExecutions` USE INDEX (PRIMARY)
    -> where ((`applicationId` > _binary'appAAAA') or (((`applicationId` = _binary'appAAAA')) AND (`id` > _binary'waxAAAA'))) and ((`applicationId` < _binary'appZZZZ') or (((`applicationId` = _binary'appZZZZ')) AND (`id` < _binary'waxZZZZ')) or ((`applicationId` = _binary'appZZZZ') and (`id` = _binary'waxZZZZ')))
    -> order by `applicationId` asc, `id` asc
    -> limit 1
    -> offset 999;
+-------------------+-------------------+
| applicationId     | id                |
+-------------------+-------------------+
...
+-------------------+-------------------+
1 row in set (0.53 sec)

Is it possible to force the use of the index when constructing the query? Is that an appropriate solution?

Langage dominant
Go
Étoiles
13.6k
Forks
1.4k
Merge moyen
9 j 19 h
PR mergées (30 j)
1

Préparer son environnement

Par où commencer

  1. Lisez l'issue en entier, puis le guide de contribution du projet.
  2. Signalez en commentaire que vous la prenez — cela évite que deux personnes fassent le même travail.
  3. Forkez le dépôt et travaillez sur une branche.
  4. Ouvrez une pull request qui référence le numéro de l'issue.

Autres issues de github/gh-ost

Toutes les issues de github/gh-ost

Issues similaires

Plus d'issues Go

Recevez les nouvelles issues par e-mail

Un résumé court des issues GitHub adaptées aux débutants.