82 seconds to extract one row in the db, java.lang.OutOfMemoryError: GC overhead limit exceeded, with large database

Question

我在数据库（本地主机）的 table 中添加了 100 000 行，从那时起我得到 this error:

java.lang.OutOfMemoryError: GC overhead limit exceeded

我通过输入 consol 解决了这个问题：

javaw -XX:-UseConcMarkSweepGC

控制台的输出是（有关上下文，请参阅下面的代码）：

2015-08-02T02:57:22.779+0200|Info: 5
2015-08-02T02:57:22.779+0200|Info: end, time taken: 82755

在数据库中提取一行需要82秒（见最后的代码）。当我的行数较少时它工作正常，所以我想知道：

为什么提取 1 行要花这么多时间？ JPA 不可能提取对象中的每一行？或者是吗？哇哦。
有办法解决这个问题吗？我的意思是在 80 秒内提取一行是非常慢的。
我真的必须输入命令 -XX:-UseConcMarkSweepGC 吗？它有什么作用？来自 the doc :

Use concurrent mark-sweep collection for the old generation. (Introduced in 1.4.1)

我的代码是这样的：

@EJB
private ThreadLookUpInterface ts;

@Schedule(hour = "*", minute = "*/1", second = "0", persistent = false)
@Override
public void makeTopThreadList() {
    System.out.println("" + ts.getThread(5).getIdthread());
}

我的服务 ejb 是这样的：

@Stateless
public class ThreadLookUpService implements ThreadLookUpInterface {

    @PersistenceContext(unitName = "my-pu")
    private EntityManager em;

    private static final String FIND_THREAD_BY_ID = "SELECT t FROM Thethread t WHERE t.idthread=:id";

    @Override
    public Thethread getThread(int threadId) {
        Query query = em.createQuery(FIND_THREAD_BY_ID);
        query.setParameter("id", threadId);
        try {
            Thethread thread = (Thethread) query.getSingleResult();
            return thread;
        } catch (NoResultException e) {
            return null;
        } catch (Exception e) {
            throw new DAOException(e);
        }
    }
}

我的实体：

@Entity
@Table(name = "thethreads")
@NamedQuery(name = "Thethread.findAll", query = "SELECT t FROM Thethread t")
public class Thethread implements Serializable {
    private static final long serialVersionUID = 1L;

    @Id
    private int idthread;

    private String content;

    @Temporal(TemporalType.TIMESTAMP)
    @Column(name = "date_posted")
    private Date datePosted;

    private int downvotes;

    @Column(name = "hot_score")
    private int hotScore;

    @Column(name = "is_pol")
    private String isPol;

    private String title;

    private String type;

    private int upvotes;

    // bi-directional many-to-one association to Category
    @OneToMany(mappedBy = "thethread", fetch = FetchType.EAGER)
    private List<Category> categories;

    // bi-directional many-to-one association to Post
    @OneToMany(mappedBy = "thethread", fetch = FetchType.EAGER)
    private List<Post> posts;

    // bi-directional many-to-one association to Category
    @ManyToOne
    @JoinColumn(name = "categories_idcategory")
    private Category category;

    // bi-directional many-to-one association to Post
    @ManyToOne
    @JoinColumn(name = "last_post")
    private Post post;

    // bi-directional many-to-one association to User
    @ManyToOne
    @JoinColumn(name = "posted_by")
    private User user;

    // bi-directional many-to-one association to ThreadVote
    @OneToMany(mappedBy = "thethread", fetch = FetchType.EAGER)
    private List<ThreadVote> threadVotes;

    public Thethread() {
    }
}

Answer 1

command -XX:-UseConcMarkSweepGC? What does it do?

设置垃圾收集器：ConcurrentMarkSweep (CMS)。 Java 堆（对象在其生命周期中所在的位置）主要分为两部分：年轻代和老年代。垃圾收集器负责在该分区上应用一些算法或策略来清理堆。 CMS 不像默认 GC 那样在完全垃圾收集期间停止应用程序线程，而是使用一个或多个后台线程定期扫描旧代并丢弃未使用的对象。这可以帮助您减少开销情况，但问题并没有消失。

extract every row in an object ? Or does it ? Just wow.Is there a way around this ? I mean extracting a single row in 80 seconds is borderline slow.

答案首先取决于您的实体模型。例如，一个常见的问题是滥用 Eager fetch 类型，这会导致 fetch/retrieve 大量不必要的对象为此执行多个查询语句。还取决于您的 JPA 实现和数据库如何解决任务，但我建议您开始查看实体模型。也许如果你 post Thethread 实体有人可以识别可能的问题。

82 seconds to extract one row in the db, java.lang.OutOfMemoryError: GC overhead limit exceeded, with large database

82 seconds to extract one row in the db, java.lang.OutOfMemoryError: GC overhead limit exceeded, with large database

java

garbage-collection

jpa

glassfish

jakarta-ee