[
https://issues.apache.org/jira/browse/IGNITE-11238?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Aleksey Plekhanov updated IGNITE-11238:
---------------------------------------
Fix Version/s: (was: 2.9)
2.10
> Possible hang on exchange
> -------------------------
>
> Key: IGNITE-11238
> URL: https://issues.apache.org/jira/browse/IGNITE-11238
> Project: Ignite
> Issue Type: Bug
> Components: general
> Reporter: Igor Seliverstov
> Priority: Critical
> Fix For: 2.10
>
>
> Currently we may hang on exchange for a while (two network timeouts) waiting
> for release a latch (see
> {{GridDhtPartitionsExchangeFuture#waitPartitionRelease releaseLatch}}) in
> case a processing topology version has not been added to discovery history
> yet but client acknowledge already received by coordinator:
> {code:java}
> [2019-02-06
> 17:43:17,009][ERROR][sys-#43%mvcc.CacheMvccPartitionedSqlCoordinatorFailoverTest0%][ExchangeLatchManager]
> Topology AffinityTopologyVersion [topVer=24, minorTopVer=0] not found in
> discovery history ; consider increasing IGNITE_DISCOVERY_HISTORY_SIZE
> property. Current value is -1
> class org.apache.ignite.IgniteException: Topology AffinityTopologyVersion
> [topVer=24, minorTopVer=0] not found in discovery history ; consider
> increasing IGNITE_DISCOVERY_HISTORY_SIZE property. Current value is -1
> at
> org.apache.ignite.internal.processors.cache.distributed.dht.preloader.latch.ExchangeLatchManager.aliveNodesForTopologyVer(ExchangeLatchManager.java:260)
> at
> org.apache.ignite.internal.processors.cache.distributed.dht.preloader.latch.ExchangeLatchManager.getLatchCoordinator(ExchangeLatchManager.java:302)
> at
> org.apache.ignite.internal.processors.cache.distributed.dht.preloader.latch.ExchangeLatchManager.processAck(ExchangeLatchManager.java:351)
> at
> org.apache.ignite.internal.processors.cache.distributed.dht.preloader.latch.ExchangeLatchManager.lambda$new$0(ExchangeLatchManager.java:121)
> at
> org.apache.ignite.internal.managers.communication.GridIoManager.invokeListener(GridIoManager.java:1561)
> at
> org.apache.ignite.internal.managers.communication.GridIoManager.processRegularMessage0(GridIoManager.java:1189)
> at
> org.apache.ignite.internal.managers.communication.GridIoManager.access$4200(GridIoManager.java:127)
> at
> org.apache.ignite.internal.managers.communication.GridIoManager$8.run(GridIoManager.java:1086)
> at
> java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1149)
> at
> java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:624)
> at java.lang.Thread.run(Thread.java:748)
> {code}
> This way the received ack won't be processed, so, we will be waiting for
> retry:
> {code:java}
> // Try to resend ack.
> releaseLatch.countDown();
> {code}
> To solve the issue we need to test whether the version is present in
> discovery history and put it into a pending map if i isn't so (see
> {{ExchangeLatchManager#pendingAcks}})
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
