maybe switch to dbcp pool to see if it behaves the same, if not can be an equals/hashcode issue
Romain Manni-Bucau @rmannibucau <https://twitter.com/rmannibucau> | Blog <https://blog-rmannibucau.rhcloud.com> | Old Blog <http://rmannibucau.wordpress.com> | Github <https://github.com/rmannibucau> | LinkedIn <https://www.linkedin.com/in/rmannibucau> | JavaEE Factory <https://javaeefactory-rmannibucau.rhcloud.com> 2017-04-26 16:45 GMT+02:00 Paul Carter-Brown < paul.carter-br...@smilecoms.com>: > Hi Romain, > > In order to better understand the commit timing, I have put some logging in > a custom MySQL ConnectionLifecycleInterceptor so that I can do a stack > trace when commit is called and see where it's called relative to the SOAP > service invocation. I also put a Thread.sleep(1000) in the commit method of > my ConnectionLifecycleInterceptor to see how it influenced the response > going back to the client. Whats very strange is that the 1000ms sleep does > result in the response taking almost 1000ms longer. I say almost, because > again when running with multiple threads, I can get a situation where the > response beats the completion of the commit. I cannot create this when the > client has only one thread. Here is a log snippet showing the scenario: > > 1) Here is our ConnectionLifecycleInterceptor getting the commit call and > logging the stack before sleeping for 1000ms. Its committing a new row with > ID 615867 > 2017/04/26 16:22:50.592 FINE > [TomEE-Exec-15][207][com.smilecoms.commons.base.lifecycle. > SmileConnectionLifecycleInterceptor.commit] > |Commit called > 2017/04/26 16:22:50.593 FINE > [TomEE-Exec-15][207][com.smilecoms.commons.base.lifecycle. > SmileConnectionLifecycleInterceptor.commit] > |Stack in commit. Sleeping now: > java.lang.Exception > at > com.smilecoms.commons.base.lifecycle.SmileConnectionLifecycleInterc > eptor.commit(SmileConnectionLifecycleInterceptor.java:32) > at com.mysql.jdbc.ConnectionImpl$2.forEach(ConnectionImpl.java: > 1593) > at com.mysql.jdbc.ConnectionImpl$2.forEach(ConnectionImpl.java: > 1592) > at com.mysql.jdbc.IterateBlock.doForAll(IterateBlock.java:61) > at com.mysql.jdbc.ConnectionImpl.commit(ConnectionImpl.java:1599) > at sun.reflect.GeneratedMethodAccessor137.invoke(Unknown Source) > at > sun.reflect.DelegatingMethodAccessorImpl.invoke( > DelegatingMethodAccessorImpl.java:43) > at java.lang.reflect.Method.invoke(Method.java:498) > at > org.apache.tomcat.jdbc.pool.ProxyConnection.invoke( > ProxyConnection.java:126) > at > org.apache.tomcat.jdbc.pool.JdbcInterceptor.invoke( > JdbcInterceptor.java:108) > at > org.apache.tomcat.jdbc.pool.DisposableConnectionFacade.invoke( > DisposableConnectionFacade.java:81) > at com.sun.proxy.$Proxy58.commit(Unknown Source) > at > org.apache.openejb.resource.jdbc.managed.local.LocalXAResource.commit( > LocalXAResource.java:131) > at > org.apache.geronimo.transaction.manager.TransactionImpl.commitResource( > TransactionImpl.java:627) > at > org.apache.geronimo.transaction.manager.TransactionImpl.commit( > TransactionImpl.java:305) > at > org.apache.geronimo.transaction.manager.TransactionManagerImpl.commit( > TransactionManagerImpl.java:252) > at > org.apache.openejb.core.transaction.JtaTransactionPolicy. > completeTransaction(JtaTransactionPolicy.java:331) > at > org.apache.openejb.core.transaction.TxRequired.commit(TxRequired.java:76) > at > org.apache.openejb.core.transaction.EjbTransactionUtil.afterInvoke( > EjbTransactionUtil.java:77) > at > org.apache.openejb.core.stateless.StatelessContainer._ > invoke(StatelessContainer.java:271) > at > org.apache.openejb.core.stateless.StatelessContainer. > invoke(StatelessContainer.java:212) > at > org.apache.openejb.server.cxf.ejb.EjbMethodInvoker. > preEjbInvoke(EjbMethodInvoker.java:125) > at > org.apache.openejb.server.cxf.ejb.EjbMethodInvoker.invoke( > EjbMethodInvoker.java:82) > at > org.apache.cxf.service.invoker.AbstractInvoker. > invoke(AbstractInvoker.java:74) > at > org.apache.cxf.interceptor.ServiceInvokerInterceptor$1. > run(ServiceInvokerInterceptor.java:59) > at > java.util.concurrent.Executors$RunnableAdapter.call(Executors.java:511) > at java.util.concurrent.FutureTask.run(FutureTask.java:266) > at > org.apache.cxf.interceptor.ServiceInvokerInterceptor$2. > run(ServiceInvokerInterceptor.java:126) > at > org.apache.cxf.workqueue.SynchronousExecutor.execute( > SynchronousExecutor.java:37) > at > org.apache.cxf.interceptor.ServiceInvokerInterceptor.handleMessage( > ServiceInvokerInterceptor.java:131) > at > org.apache.cxf.phase.PhaseInterceptorChain.doIntercept( > PhaseInterceptorChain.java:308) > at > org.apache.cxf.transport.ChainInitiationObserver.onMessage( > ChainInitiationObserver.java:121) > at > org.apache.cxf.transport.http.AbstractHTTPDestination.invoke( > AbstractHTTPDestination.java:262) > at > org.apache.openejb.server.cxf.CxfWsContainer.onMessage( > CxfWsContainer.java:85) > at > org.apache.openejb.server.webservices.WsServlet.service(WsServlet.java:98) > at > org.apache.catalina.core.ApplicationFilterChain.internalDoFilter( > ApplicationFilterChain.java:230) > at > org.apache.catalina.core.ApplicationFilterChain.doFilter( > ApplicationFilterChain.java:165) > at > org.apache.catalina.core.StandardWrapperValve.invoke( > StandardWrapperValve.java:198) > at > org.apache.catalina.core.StandardContextValve.invoke( > StandardContextValve.java:96) > at > org.apache.catalina.core.StandardHostValve.invoke( > StandardHostValve.java:140) > at > org.apache.catalina.valves.ErrorReportValve.invoke( > ErrorReportValve.java:79) > at > org.apache.tomee.catalina.OpenEJBSecurityListener$RequestCapturer.invoke( > OpenEJBSecurityListener.java:97) > at > org.apache.catalina.core.StandardEngineValve.invoke( > StandardEngineValve.java:87) > at > org.apache.catalina.connector.CoyoteAdapter.service( > CoyoteAdapter.java:349) > at > org.apache.coyote.http11.Http11Processor.service(Http11Processor.java:783) > at > org.apache.coyote.AbstractProcessorLight.process( > AbstractProcessorLight.java:66) > at > org.apache.coyote.AbstractProtocol$ConnectionHandler.process( > AbstractProtocol.java:798) > at > org.apache.tomcat.util.net.NioEndpoint$SocketProcessor. > doRun(NioEndpoint.java:1434) > at > org.apache.tomcat.util.net.SocketProcessorBase.run( > SocketProcessorBase.java:49) > at > java.util.concurrent.ThreadPoolExecutor.runWorker( > ThreadPoolExecutor.java:1142) > at > java.util.concurrent.ThreadPoolExecutor$Worker.run( > ThreadPoolExecutor.java:617) > at > org.apache.tomcat.util.threads.TaskThread$WrappingRunnable.run( > TaskThread.java:61) > at java.lang.Thread.run(Thread.java:745) > > 2) Now here is the clients request being processed in our EJB looking for > the row that was created above but is not finished committing yet. The row > with ID 615867 is not found. > 2017/04/26 16:22:51.588 FINE > [TomEE-Exec-10][202][com.smilecoms.im.IdentityManager.getCustomers] > |Looking up a single customer with ID [615867] > 3) Here is the sleep having finished. > 2017/04/26 16:22:51.594 FINE > [TomEE-Exec-15][207][com.smilecoms.commons.base.lifecycle. > SmileConnectionLifecycleInterceptor.commit] > |Commit sleep finished > > I would expect that only after 2017/04/26 16:22:51.594 would the client > receive the SOAP response to the insert but clearly that is not the case. > It is slowed down by the sleep but there still seems to be some race > condition where it somehow can still beat the commit returning. If the > commit were called after sending the response or in parallel to doing so > then I would have expected the 1000ms sleep to not impact the response time > of the service. Very very strange.... > > Here is the code snippet of my SmileConnectionLifecycleInterceptor: > > public class SmileConnectionLifecycleInterceptor implements > ConnectionLifecycleInterceptor { > > private static final Logger log = > LoggerFactory.getLogger(SmileConnectionLifecycleInterceptor.class); > > > @Override > public boolean commit() throws SQLException { > log.debug("Commit called"); > log.debug("Stack in commit. Sleeping now: ", new Exception()); > try { > Thread.sleep(1000); > } catch (Exception e) { > > } > log.debug("Commit sleep finished"); > return true; > } > ... > > > Any thoughts? > > Paul > > On 26 April 2017 at 13:50, Paul Carter-Brown < > paul.carter-br...@smilecoms.com> wrote: > > > Thought so. Let me try and create a reproducable test case and let you > > know what i find > > > > Paul > > > > Sent from my phone. Message may be brief and contain typos. > > > > > > On 26 Apr 2017 13:35, "Romain Manni-Bucau" <rmannibu...@gmail.com> > wrote: > > > >> Hi > >> > >> Shouldn't be possible since it would assume this flow: > >> > >> thread1. start tx | thread2. - > >> thread1. create the entry | thread2. - > >> thread1. serialize the entry | thread2. - > >> thread1. commit tx | thread2. - find the entry > >> thread1. | thread2. - fails > >> > >> The commit will happen > >> in org.apache.openejb.server.cxf.ejb.EjbMethodInvoker#invoke where the > >> serialization happens after > >> in org.apache.cxf.jaxb.io.DataWriterImpl#write(java.lang.Object, > >> org.apache.cxf.service.model.MessagePartInfo, T) > >> > >> Only case it should be possible to happen is if you wrapped jaxws > >> servlet/layer with a custom transaction management (through a filter for > >> instance) which would start/commit outside default JAXWS handling. > >> > >> > >> Romain Manni-Bucau > >> @rmannibucau <https://twitter.com/rmannibucau> | Blog > >> <https://blog-rmannibucau.rhcloud.com> | Old Blog > >> <http://rmannibucau.wordpress.com> | Github < > >> https://github.com/rmannibucau> | > >> LinkedIn <https://www.linkedin.com/in/rmannibucau> | JavaEE Factory > >> <https://javaeefactory-rmannibucau.rhcloud.com> > >> > >> 2017-04-26 12:59 GMT+02:00 Paul Carter-Brown < > >> paul.carter-br...@smilecoms.com>: > >> > >> > Hi > >> > > >> > I have a call flow as follows in tomee: > >> > > >> > 1. Client calls jaxws service that adds a row into mysql table. EJB is > >> > using JTA with eclipselink JPA entity bean. Seevice returns the > primary > >> key > >> > of the row just added. > >> > 2. Client uses returned soap message that contains the primary key and > >> > calls a second service that does a select for the row in question. > >> > > >> > The service in 1 and 2 is in the same tomee instance and caching in > jpa > >> is > >> > disabled. Tx isolation isread committed. > >> > > >> > Under high load and concurrency doing about 1000 tps i get about a 20% > >> > failure rate where the service at stage 2 cannot find the row inserted > >> at > >> > stage 1. The row is inserted at some stage as i do see it in the db. > >> Its as > >> > though service 1 is returning prior to the database having committed > the > >> > data. My understanding is that this should not be possible as it would > >> > create all sorts of inconsistencies. Is what i am observing expected > >> > behaviour? > >> > > >> > Paul > >> > > >> > Sent from my phone. Message may be brief and contain typos. > >> > > >> > -- > >> > > >> > > >> > This email is subject to the disclaimer of Smile Communications at > >> > http://www.smilecoms.com/home/email-disclaimer/ < > >> http://www.smilecoms.com/ > >> > disclaimer> > >> > > >> > > >> > > > > > -- > > *Paul Carter-Brown* > > *Group Chief Information Officer* > > *Smile Communications Pty (Ltd) * > Smile +234 (0) 702 000 1234 > Mobile +27 (0) 83 4427 179 > Skype PaulC-B > paul.carter-br...@smilecoms.com > www.smilecoms.com > > -- > > > This email is subject to the disclaimer of Smile Communications at > http://www.smilecoms.com/home/email-disclaimer/ <http://www.smilecoms.com/ > disclaimer> > >
