如何调试这个 Hibernate/JPA 慢查询问题？

Question

我有一个 Hibernate/JPA 查询，它基于数据库中的索引字段执行简单查询。该表相当大，大约有 2800 万条记录，但是当直接使用数据库时，查询会立即返回，而使用 Hibernate+JPA 则需要 20 多秒。数据库是 PostgreSQL 9.0.4。

这是包含相关查询的日志摘录，从时间戳中您可以看到在发出查询和打开 ResultSet 之间存在很长的延迟。

2011-09-28 06:26:38,019 [http-thread-pool-8080(5)] DEBUG org.hibernate.jdbc.AbstractBatcher - about to open PreparedStatement (open PreparedStatements: 0, globally: 0)
2011-09-28 06:26:38,019 [http-thread-pool-8080(5)] DEBUG org.hibernate.jdbc.ConnectionManager - opening JDBC connection
2011-09-28 06:26:38,020 [http-thread-pool-8080(5)] DEBUG org.hibernate.SQL - select pafaddress0_.address_id as address1_0_, pafaddress0_.admin1 as admin2_0_, pafaddress0_.admin2 as admin3_0_, pafaddress0_.admin3 as admin4_0_, pafaddress0_.barcode as barcode0_, pafaddress0_.building_name as building6_0_, pafaddress0_.building_number as building7_0_, pafaddress0_.country_name as country8_0_, pafaddress0_.dependent_locality as dependent9_0_, pafaddress0_.dependent_street as dependent10_0_, pafaddress0_.dept_name as dept11_0_, pafaddress0_.double_dep_locality as double12_0_, pafaddress0_.dps as dps0_, pafaddress0_.mailsort as mailsort0_, pafaddress0_.org_name as org15_0_, pafaddress0_.pca_id as pca16_0_, pafaddress0_.pobox as pobox0_, pafaddress0_.postcode as postcode0_, pafaddress0_.postcode1 as postcode19_0_, pafaddress0_.postcode_type as postcode20_0_, pafaddress0_.street as street0_, pafaddress0_.su_org_indicator as su22_0_, pafaddress0_.subbuilding_name as subbuil23_0_, pafaddress0_.posttown as posttown0_, pafaddress0_.trad_county as trad25_0_, pafaddress0_.type as type0_, pafaddress0_.udprn as udprn0_ from paf pafaddress0_ where pafaddress0_.pca_id=? limit ?
2011-09-28 06:27:01,897 [http-thread-pool-8080(5)] DEBUG org.hibernate.jdbc.AbstractBatcher - about to open ResultSet (open ResultSets: 0, globally: 0)
2011-09-28 06:27:01,900 [http-thread-pool-8080(5)] DEBUG org.hibernate.jdbc.AbstractBatcher - about to close ResultSet (open ResultSets: 1, globally: 1)
2011-09-28 06:27:01,900 [http-thread-pool-8080(5)] DEBUG org.hibernate.jdbc.AbstractBatcher - about to close PreparedStatement (open PreparedStatements: 1, globally: 1)
2011-09-28 06:27:01,900 [http-thread-pool-8080(5)] DEBUG org.hibernate.jdbc.ConnectionManager - aggressively releasing JDBC connection
201

我的实体没有复杂的关联，日志中也没有任何迹象表明在这段漫长的延迟期间发生了什么。

@Entity
@Table(name = "paf")
@NamedQueries({
        @NamedQuery(name = "PafAddress.findByPcaId", query = "SELECT a FROM PafAddress a where a.pcaId=:pcaId ")
})
public class PafAddress implements Serializable, Identifiable<Long> {
    private static final long serialVersionUID = -5831210090000717285L;

    @Id
    @Column(name = "address_id")
    private Long id;

    private String postcode;

    @Column(name = "posttown")
    private String town;

    @Column(name = "dependent_locality")
    private String dependentLocality;

    @Column(name = "double_dep_locality")
    private String doubleDepLocality;

    @Column(name = "street")
    private String street;

    @Column(name = "dependent_street")
    private String dependentStreet;

    @Column(name = "building_number")
    private String buildingNumber;

    @Column(name = "building_name")
    private String buildingName;

    @Column(name = "subbuilding_name")
    private String subbuildingName;

    @Column(name = "pobox")
    private String pobox;

    @Column(name = "dept_name")
    private String deptName;

    @Column(name = "org_name")
    private String orgName;

    @Column(name = "udprn")
    private String udprn;

    @Column(name = "postcode_type")
    private String postcodeType;

    @Column(name = "su_org_indicator")
    private String suOrgIndicator;

    @Column(name = "dps")
    private String dps;

    @Column(name = "postcode1")
    private String postcode1;

    @Column(name = "admin1")
    private String admin1;

    @Column(name = "admin2")
    private String admin2;

    @Column(name = "admin3")
    private String admin3;

    @Column(name = "trad_county")
    private String tradCounty;

    @Column(name = "country_name")
    private String countryName;

    @Column(name = "mailsort")
    private String mailsort;

    @Column(name = "barcode")
    private String barcode;

    @Column(name = "type")
    private String type;

    @Column(name = "pca_id")
    private Double pcaId;

    public PafAddress() {
    }

关于为什么此查询执行如此糟糕或我如何确定原因的任何想法？

Answer 1

我发现了问题：

我在 PostgreSQL 中为耗时 >1 秒的查询启用了 auto_explain，这使我能够在日志中查看查询计划。 见http://www.postgresql.org/docs/current/static/auto-explain.html

对于 Hibernate/JPA 查询：

Limit  (cost=0.00..13.12 rows=2 width=469)
  ->  Seq Scan on paf pafaddress0_  (cost=0.00..943043.74 rows=143753 width=469)
        Filter: ((pca_id)::double precision = 7912139::double precision)

原生数据库查询：

                                          QUERY PLAN                                              
 Limit  (cost=0.00..7.03 rows=2 width=469)
   ->  Index Scan using paf_pca_idx on paf pafaddress0_  (cost=0.00..505649.83 rows=143753 width=469)
         Index Cond: (pca_id = 7912139::numeric)

这很慢，因为正在执行顺序扫描而不是使用索引，这似乎是由于我的实体和数据库字段之间的类型不匹配造成的。实体属性是“Double”，而数据库类型是“numeric”，查询说明显示了数据库字段的转换为 double double precision。

我将我的数据库字段类型更改为double precision，查询现在按预期执行。

Answer 2

如果sql相同，瓶颈一定是为每条记录创建单独的java对象的过程。我假设它返回大量记录，您可以尝试限制这个数量。