activemq-users mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From michael.andre.pea...@me.com.INVALID
Subject Re: AMQ 224038 on Last Value Queue
Date Fri, 05 Jul 2019 07:27:15 GMT
Have you done the same test with lvq and non destructive false (default) where the consumer
receives the message but doesnt ack it. Then restart broker. 




If the message is lost there it would suggest issue in lvq or persistence. 




Remember you need to set the queue to durable if you want it to survive a restart. 




I did a quick test myself and from what i can tell if the queue is non durable i get the experiece
you are describing which would be expected as queue is not durable. But if i set the queue
durable i dont. 




So far from tests ive done things are behaving expected per settings.






Get Outlook for Android







On Fri, Jul 5, 2019 at 8:06 AM +0100, <michael.andre.pearce@me.com.INVALID> wrote:










Note non destructive, is regardless of the queue being lvq or normal. 




Though primary case (but not only) for it is in conjunction with lvq.




Get Outlook for Android







On Fri, Jul 5, 2019 at 8:03 AM +0100,  wrote:










When non destructive is true the consumer ack ignored.




When non destructive is false (default) the consumer ack is honoured.




There is test cases for this and is used in real use cases.




Get Outlook for Android







On Fri, Jul 5, 2019 at 7:30 AM +0100, "yw yw"  wrote:










Hi,
Just to clarify,  are consumers meant to be able to remove the message for
lvq when non destructive is true?

I took a test: for the lvq of which non destructive is true, the producer
sent a message and consumer received it. After broker restarted, the
message was removed.

 于2019年7月5日周五 下午1:33写道:

> For lvq when non destructive is false consumers are meant to be able to
> remove the message by acking.
>
>
>
>
> Use case there is coalesced price updates where you just care about
> lastest value, but the updates are faster than a consumer may deal with.
>
>
>
>
> Behaviour is as expected.
>
>
>
>
> Get Outlook for Android
>
>
>
>
>
>
>
> On Fri, Jul 5, 2019 at 2:53 AM +0100, "yw yw"  wrote:
>
>
>
>
>
>
>
>
>
>
> Yes, the last value always wins.
>
> The document says "Another common pattern is to have queue "browsers" which
> send all messages to the browser, but do not prevent other consumers from
> receiving the messages, and do not remove them from the queue when the
> browser is done with them. Such a browser is an instance of a
> "non-destructive" consumer." The fact is we don't remove them in memory but
> we append records in journal. When the consumer acks the message and
> reference count of message is decremented to zero, the ack record and
> delete message record are written into journal. If broker restarts, the
> last value is lost. Not sure it's what we expect?
>
>
> michael.andre.pearce  于2019年7月4日周四
> 下午11:30写道:
>
> > Non destructive is so a consumer doesnt ack the message. Essentially
> > meaning the last value is always kept in the lvq.When a new messages
> > replaces and old then it needs the old one is acked to it is removed,
> this
> > is the point.Sent from my Samsung Galaxy smartphone.
> > -------- Original message --------From: yw yw  Date:
> > 04/07/2019  08:42  (GMT+00:00) To: users@activemq.apache.org Subject:
> Re:
> > AMQ 224038 on Last Value Queue Hi,We have encountered the same problems
> > these days.Right now the JMSNonDestructiveTest passes successfully bcs
> > persistence isdisabled which means there are no journal operations. If we
> > enablepersistence, the test fails in testNonDestructiveLVQTombstone().The
> > reproduce step is the same with what Matt said:1. First send a message.2.
> > Then receive the message. The key point is here: We ack this message
> > anddelete the message so the record is removed in "records" map.3. Last
> > send another message. The exception occurs
> >
> below:[Thread-13(ActiveMQ-server-org.apache.activemq.artemis.core.server.impl.ActiveMQServerImpl$6@77a57272
> )]15:01:06,990
> > ERROR [org.apache.activemq.artemis.core.server] AMQ224038:Failed to ack
> old
> > reference: java.lang.IllegalStateException: Cannot findadd info 446 on
> > compactor or current
> >
> recordsatorg.apache.activemq.artemis.core.journal.impl.JournalImpl.checkKnownRecordID(JournalImpl.java:1081)[:]atorg.apache.activemq.artemis.core.journal.impl.JournalImpl.appendUpdateRecord(JournalImpl.java:888)[:]atorg.apache.activemq.artemis.core.journal.Journal.appendUpdateRecord(Journal.java:98)[:]atorg.apache.activemq.artemis.core.persistence.impl.journal.AbstractJournalStorageManager.storeAcknowledge(AbstractJournalStorageManager.java:425)[:]atorg.apache.activemq.artemis.core.server.impl.QueueImpl.acknowledge(QueueImpl.java:1539)[:]atorg.apache.activemq.artemis.core.server.impl.LastValueQueue.acknowledge(LastValueQueue.java:193)[:]atorg.apache.activemq.artemis.core.server.impl.MessageReferenceImpl.acknowledge(MessageReferenceImpl.java:235)[:]atorg.apache.activemq.artemis.core.server.impl.LastValueQueue.replaceLVQMessage(LastValueQueue.java:172)[:]atorg.apache.activemq.artemis.core.server.impl.LastValueQueue.addTail(LastValueQueue.java:107)[:]atorg.apache.activemq.artemis.core.server.impl.RoutingContextImpl.internalprocessReferences(RoutingContextImpl.java:164)[:]atorg.apache.activemq.artemis.core.server.impl.RoutingContextImpl.processReferences(RoutingContextImpl.java:159)[:]atorg.apache.activemq.artemis.core.postoffice.impl.PostOfficeImpl$2.done(PostOfficeImpl.java:1378)[:]atorg.apache.activemq.artemis.core.persistence.impl.journal.OperationContextImpl$1.run(OperationContextImpl.java:244)[:]atorg.apache.activemq.artemis.utils.actors.OrderedExecutor.doTask(OrderedExecutor.java:42)[:]atorg.apache.activemq.artemis.utils.actors.OrderedExecutor.doTask(OrderedExecutor.java:31)[:]atorg.apache.activemq.artemis.utils.actors.ProcessorBase.executePendingTasks(ProcessorBase.java:66)[:]atjava.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)[rt.jar:1.8.0_102]atjava.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)[rt.jar:1.8.0_102]atorg.apache.activemq.artemis.utils.ActiveMQThreadFactory$1.run(ActiveMQThreadFactory.java:118)[:]I'm
> > confused with the current lastValueQueue design. Why do we acknowledgethe
> > message(i mean in normal ack reason) if the queue is non destructive?What
> > if we acknowledged messages and messages were deleted in the journal,then
> > broker restarted? I assume these messages would be lost, seems againstthe
> > non destructive principal? It seems like we don't need to
> reallyacknowledge
> > messages except for KILLED/EXPIRED/ REPLACED reasons for thenon
> destructive
> > queue?Justin Bertram  于2019年7月4日周四 上午11:27写道:> There
> > are a couple of tests in the test suite which use both last-value and>
> > non-destructive queue attributes (e.g. in>
> >
> org.apache.activemq.artemis.tests.integration.amqp.JMSNonDestructiveTest).>
> > These work without issue. I also took the code you pasted and tried to>
> > reproduce the failure, but everything appears to be working as expected.>
> > For what it's worth, I'm testing on the tip of the master branch (i.e.>
> > 2.10.0-SNAPSHOT).>> Would it be possible for you to provide more details
> > about how to reproduce> this failure? Perhaps some simple modifications
> to
> > the last-value-queue> example distributed with the broker would
> suffice.>>>
> > Justin>> On Wed, Jul 3, 2019 at 4:44 PM mschmeiser
> > wrote:>> > Hello> >> > I am getting an error "AMQ224039: Failed
to ack
> old
> > reference:> > java.lang.IllegalStateException: Cannot find add info 5698
> on
> > compactor> or> > current records" after every publish of a message to a
> > Last Value Queue> > except the first time.> >> > I am on ActiveMQ
Artemis
> > 2.8.1. I know there's a more recent version but> > checked the release
> > notes and none of the bug fixes looked like they'd be> > relevant to
> this.>
> > >> > I'm not working with a deployment-ready system yet, this is very
> > early> > stage> > prototyping. I have a single broker with the basic
> > configuration, I'm> > basically running things 'out of the box'. The java
> > code that I'm using> > looks something like:> >> > ***> >>
>
> ActiveMQQueue
> > destination = ActiveMQQueue.createQueue(subject +> >
> > "?last-value-key=LVK&non-destructive=true");> >> > MessageProducer
> producer
> > = session.createProducer(destination);> >> > TextMessage message =
> > session.createTextMessage("test object");> >
> > message.setStringProperty("LVK", "foo");> >> > Thread.sleep(500);>
>> >
> > session.createConsumer(destination, "LVK = 'foo'").setMessageListener(m>
> > ->> > processMessage(m));> >> > Thread.sleep(500);> >>
> message =
> > session.createTextMessage("test object 2");> >
> > message.setStringProperty("LVK", "foo");> >> > producer.send(message);>
> >>
> > > Thread.sleep(500);> >> > session.createConsumer(destination, "LVK
=
> > 'foo'").setMessageListener(m> ->> > processMessage(m));> >>
> ***> >> >
> The
> > first time this code is run, I only get the error when the message is> >
> > sent the second time. After that, the error happens twice per run. If I>
> > go> > into the broker console and purge the queue, then the first message
> > send> > works fine, error is thrown every subsequent time.> >> >
The
> stack
> > trace (I'm not able to copy/paste from my development> > environment)>
>
> > looks like:> >> > JournalImpl.checkKnownRecordID(JournalImpl.java:1074)>
> >
> > JournalImpl.appendUpdateRecord(JournalImpl.java:881)> >
> > Journal.appendUpdateRecord(Journal.java:98)> >> >>
> >
> AbstractJournalStorageManager.storeAcknowledge(AbstractJournalStorageManager.java:425)>
> > > QueueImpl.acknowledge(QueueImpl.java:1533)> >
> > LastValueQueue.acknowledge(MessageReferenceImpl.java:235)> >
> > MessageReferenceImpl.acknowledge(MessageReferenceImpl.java:235)> >
> > LastValueQueue.replaceLVQMessage(LastValueQueue.java:172)> >
> > LastValueQueue.addTail(LastValueQueue.java:107)> > ...> >> > Tracing
it
> > back that far, it looks like the failed acknowledge shouldn't> > cause
> > major problems for the functionality of the system. The code is all> >
> > executing as you'd expect it to and the queue on the broker always has>
> > the> > expected message in it. The error is still a concern though simply
> > as a> > matter of performance - scalability is a concern for me and the
> > overhead> of> > logging a bunch of errors would be a problem.> >>
> Let
> me
> > know if there's any other information I can provide that could> help> >
> > diagnose this issue.> >> > Thanks,> > Matt> >> >>
>> > --> > Sent from:>
> >
> > http://activemq.2283324.n4.nabble.com/ActiveMQ-User-f2341805.html> >>
>
>
>
>
>
>
















Mime
  • Unnamed multipart/alternative (inline, None, 0 bytes)
View raw message