Hibernate的线程问题 [英] Threading issues with Hibernate
问题描述
我遇到了一个奇怪的行为与Hibernate。我已经把我的头撞在墙上了一段时间,并将奖励任何答案,导致一个+100奖金的解决方案。
我有一个JAX -RS(Jersey)REST服务器,带有一个过滤器,用于关联每个请求的一个Hibernate会话。
在一个请求中,客户端POST存储在数据库中的一些数据一个会话(和一个事务)。在后续调用中,客户端尝试获取此实体,但Hibernate找不到它。
一些观察结果:
-
如果我同时运行多个客户端,我只能重现这个。 )
-
我可在实体ID中查看实体ID。数据库,如果我重新启动服务器,实体被Hibernate找到它应该。
-
如果我使用线程池大小的错误不会发生
这是代码,有一些日志记录:
chargeables.setId(new SecureRandom()。nextLong());
System.out.printf(%s,session:%s [%s]%n,
Thread.currentThread(),
System.identityHashCode ,
session.beginTransaction());
session.beginTransaction();
System.out.printf(%s,session:%s [%s]%n,
Thread.currentThread(),
系统。 identityHashCode(session),
session.save(id =+ chargeables.getId()+));
session.save(chargeables);
System.out.printf(%s,session:%s [%s]%n,
Thread.currentThread(),
系统。 identityHashCode(session),
session.getTransaction()。commit());
session.getTransaction()。commit();
获取实体的代码:
System.out.printf(%s,session:%s [%s]%n,
Thread.currentThread(),
System.identityHashCode会话),
session.get(+ id +));
Chargeables entity =(Chargables)session.get(Chargables.class,id);
if(entity == null)
System.out.printf(%s,session:%s [%s]%n,
Thread.currentThread ,
System.identityHashCode(session),
ENTITY NOT FOUND!);
现在这里是一个摘录的结果日志(与一些额外的打开/关闭会话输出):
线程[Grizzly(5),5,main],session:2041842357 [factory.openSession()]
[Grizzly(5),5,main],session:2041842357 [session.save(id = 7939229356942262438)] [session.beginTransaction()]
线程[Grizzly(5),5,main],session:2041842357 [session.getTransaction()。commit()会话:1717445911 [factory.openSession()]
线程[Grizzly(7),5,main] ),session:1717445911 [session.get(7939229356942262438)]
线程[Grizzly(7),5,main],session:1717445911 [ENTITY NOT FOUND!]
Thread [Grizzly (7),5,main],session:1717445911 [session.close()]
地球到达 ENTITY未找到
Hibernate版本:4.1.9.Final├
MySQL verison:14.14 Distrib 5.5.29
映射文件费用
:
<?xml version =1.0? >
<!DOCTYPE hibernate-mapping PUBLIC
- // Hibernate / Hibernate Mapping DTD 3.0 // EN
http://www.hibernate.org/dtd/hibernate-mapping -3.0.dtd>
< hibernate-mapping
default-cascade =all
package =se.package.common.chargables
default-lazy =false >
< class name =Chargables>
< id name =idtype =long>
< generator class =assigned/>
< / id>
< property name =startTimeStamp/>
< property name =endTimeStamp/>
< list name =feeables>
< key column =chargeableId/>
< list-index column =pos/>
< many-to-many class =Chargeable/>
< / list>
< / class>
< class name =Chargeable>
< id column =idtype =long>
< generator class =native/>
< / id>
< discriminator />
< property name =timestamp/>
< / class>
< subclass name =DataTransferextends =Chargeable>
< property name =bytesSent/>
< property name =bytesReceived/>
< / subclass>
< subclass name =TelephonyChargeableextends =Chargeable>
< many-to-one name =num/>
< / subclass>
< subclass name =Callextends =TelephonyChargeable>
< property name =duration/>
< / subclass>
< subclass name =OutgoingCallextends =Call/>
< subclass name =IncomingCallextends =Call/>
< subclass name =Messageextends =TelephonyChargeable/>
< subclass name =Smsextend =Message/>
< subclass name =IncomingSmsextends =Sms/>
< subclass name =OutgoingSmsextends =Sms/>
< subclass name =Mmsextends =Message/>
< subclass name =IncomingMmsextends =Mms/>
< subclass name =OutgoingMmsextends =Mms/>
< / hibernate-mapping>
如果您可以在数据库中看到提交不能是MySQL级别的隔离问题:控制台操作也是客户端操作,运行它们时不会获得特殊权限,因此它们符合您选择的隔离策略。
寻找解决方案,我发现:实际上有2级缓存实现。也许你的hibernate安装捆绑了一个缓存机制?
第一级在每个会话的基础上工作,这与你在一个代码上运行时的行为是一致的线程,似乎恰好是默认激活。我不是Hibernate精通,我不能说肯定,但我的意见是,你的设置有这个第一级的缓存设置为只读模式,而你希望它处于读写模式。 p>
I'm experiencing a strange behaviour with Hibernate. I've been banging my head against the wall for a while now, and will award any answer which leads to a solution with a +100 bounty.
I have a JAX-RS (Jersey) REST server, with a filter that associates one Hibernate-session per request.
In one request a client POSTs some data which is stored in the database using one session (and one transaction). In a subsequent call, the client tries to GET this entity, but Hibernate can't find it.
Some observations:
I can only reproduce this if I run multiple simultaneous clients. I've never managed to reproduce it by running one client at a time.)
I can see the entity ID in the database, and if I restart the server, the entity is found by Hibernate as it should.
The error does not occur if I use a thread pool of size 1 (regardless of how many clients I run simultaneously).
Here's the code, with some logging:
chargeables.setId(new SecureRandom().nextLong());
System.out.printf("%s, session: %s [%s]%n",
Thread.currentThread(),
System.identityHashCode(session),
"session.beginTransaction()");
session.beginTransaction();
System.out.printf("%s, session: %s [%s]%n",
Thread.currentThread(),
System.identityHashCode(session),
"session.save(id = "+chargeables.getId()+")");
session.save(chargeables);
System.out.printf("%s, session: %s [%s]%n",
Thread.currentThread(),
System.identityHashCode(session),
"session.getTransaction().commit()");
session.getTransaction().commit();
The code for getting the entity:
System.out.printf("%s, session: %s [%s]%n",
Thread.currentThread(),
System.identityHashCode(session),
"session.get("+id+")");
Chargeables entity = (Chargeables) session.get(Chargeables.class, id);
if (entity == null)
System.out.printf("%s, session: %s [%s]%n",
Thread.currentThread(),
System.identityHashCode(session),
"ENTITY NOT FOUND!");
Now here is an excerpt of the resulting log (with some additional open/close session output):
Thread[Grizzly(5),5,main], session: 2041842357 [factory.openSession()]
Thread[Grizzly(5),5,main], session: 2041842357 [session.beginTransaction()]
Thread[Grizzly(5),5,main], session: 2041842357 [session.save(id = 7939229356942262438)]
Thread[Grizzly(5),5,main], session: 2041842357 [session.getTransaction().commit()]
Thread[Grizzly(5),5,main], session: 2041842357 [session.close()]
[...]
Thread[Grizzly(7),5,main], session: 1717445911 [factory.openSession()]
Thread[Grizzly(7),5,main], session: 1717445911 [session.get(7939229356942262438)]
Thread[Grizzly(7),5,main], session: 1717445911 [ENTITY NOT FOUND!]
Thread[Grizzly(7),5,main], session: 1717445911 [session.close()]
Why on earth do I reach ENTITY NOT FOUND!
?
Hibernate version: 4.1.9.Final
MySQL verison: 14.14 Distrib 5.5.29
Mapping file for Chargeables
:
<?xml version="1.0"?>
<!DOCTYPE hibernate-mapping PUBLIC
"-//Hibernate/Hibernate Mapping DTD 3.0//EN"
"http://www.hibernate.org/dtd/hibernate-mapping-3.0.dtd">
<hibernate-mapping
default-cascade="all"
package="se.package.common.chargeables"
default-lazy="false">
<class name="Chargeables">
<id name="id" type="long">
<generator class="assigned"/>
</id>
<property name="startTimeStamp" />
<property name="endTimeStamp" />
<list name="chargeables">
<key column="chargeableId" />
<list-index column="pos" />
<many-to-many class="Chargeable"/>
</list>
</class>
<class name="Chargeable">
<id column="id" type="long">
<generator class="native"/>
</id>
<discriminator />
<property name="timestamp" />
</class>
<subclass name="DataTransfer" extends="Chargeable">
<property name="bytesSent" />
<property name="bytesReceived" />
</subclass>
<subclass name="TelephonyChargeable" extends="Chargeable">
<many-to-one name="num" />
</subclass>
<subclass name="Call" extends="TelephonyChargeable">
<property name="duration" />
</subclass>
<subclass name="OutgoingCall" extends="Call" />
<subclass name="IncomingCall" extends="Call" />
<subclass name="Message" extends="TelephonyChargeable" />
<subclass name="Sms" extends="Message" />
<subclass name="IncomingSms" extends="Sms" />
<subclass name="OutgoingSms" extends="Sms" />
<subclass name="Mms" extends="Message" />
<subclass name="IncomingMms" extends="Mms" />
<subclass name="OutgoingMms" extends="Mms" />
</hibernate-mapping>
If you can see the commit in the database, it cannot be an isolation issue at the MySQL level: console operations are also client operations, you don't get special powers when running them, so they conform to the isolation policy you selected.
Looking out for solution, I discovered that Hibernate provide some facilities to cache DB results: there are actually 2 levels of cache implemented there. Perhaps your hibernate installation comes bundled with a cache mechanism?
The first level works on a per session basis, which is consistent with the behaviour you observe when running the code on one thread, and seemingly happens to be activated by default. I am not Hibernate proficient, I can't say that for sure, but my opinion is that your setup has this first level of cache set in read-only mode, while you would like it to be in read-write mode.
这篇关于Hibernate的线程问题的文章就介绍到这了,希望我们推荐的答案对大家有所帮助,也希望大家多多支持IT屋!