You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@ignite.apache.org by "Tomasz Grygo (Jira)" <ji...@apache.org> on 2020/06/01 17:01:00 UTC
[jira] [Commented] (IGNITE-13093) Unidentified Apache Ignite worker
blocked when inserting large amount of records to the persistent storage
[ https://issues.apache.org/jira/browse/IGNITE-13093?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17121169#comment-17121169 ]
Tomasz Grygo commented on IGNITE-13093:
---------------------------------------
I tried your suggestions and exception did not happen :). I changed maxSize to 4GB ,-Xmx30g and adjusted config settings as you proposed.
But why did it help exactly? What criteria should I use for sizing maxCache and heap? What value should I set if i have 2x or 4x the data?
I understood that maxSize is the maximum size of cache in memory. I have 130 caches and most of them will not fit in memory anyways so didn't want to unnecessarily stress the garbage collector.
Also, i have some additional questions about setting up Ignite. Is here a good place to ask them or is there some other, better place?
> Unidentified Apache Ignite worker blocked when inserting large amount of records to the persistent storage
> ----------------------------------------------------------------------------------------------------------
>
> Key: IGNITE-13093
> URL: https://issues.apache.org/jira/browse/IGNITE-13093
> Project: Ignite
> Issue Type: Bug
> Components: cache
> Affects Versions: 2.8.1
> Environment: Java 1.8.0_231
> Apache Ignite 2.8.1
> Windows 10, 64G memory
> Java settings
> -Xms1024m -Xmx50g -Xss1024m
> -Xverify:none
> -server
> -DIGNITE_QUIET=true
> -XX:+UseG1GC
> -XX:+DisableExplicitGC
> -Djava.net.preferIPv4Stack=true
> -XX:+AlwaysPreTouch
> -XX:+ScavengeBeforeFullGC
> -XX:+AggressiveOpts
> Reporter: Tomasz Grygo
> Priority: Blocker
> Attachments: 20_05_29__14_51.out.log, PureIgniteDynamicRowStorage.java, PureIgniteUtils.java, ignite.xml, log4j2.xml, thread_dump.txt
>
>
> I'm looking at Apache Ignite to use as a fast database. Performance is very important, I need to build it as fast as possible with resources available. First I copy all (450M) records from my original test database to Ignite caches through IgniteDataStreams using PK as a key. Database does not fit in memory so I have disk persistence enabled and eviction disabled. Data is inserted in parallel using 8 threads. I have only one but fairly powerful Windows PC doing all the work, no separate Ignite cluster. I'm not interested in cache recovery so WAL is disabled. Everything goes well until I hit around 310 million entries (2 hours of work). At this point Ignite starts to choke, inserts slow down and then stop with exceptions. Exception is triggered by systemWorkerBlockedTimeout setting set to 5 minutes. Extending this time does not help at all. Based on heap dump I tried adding -DIGNITE_PAGES_LIST_DISABLE_ONHEAP_CACHING=true and it failed slightly later but still could not finish the job. I read the performance guides and I tried tweaking other Ignite settings too but didn't see any impact. How can if find which worker is being blocked and why?
> 2020-05-27 21:54:26,176 [Storage2 ] [ERROR] - DTR_0030 worker Storage2 had error: FATAL ERROR java.lang.IllegalStateException: Data streamer has been closed.
> java.lang.IllegalStateException: Data streamer has been closed.
> at org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl.closedException(DataStreamerImpl.java:1095)
> at org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl.lock(DataStreamerImpl.java:446)
> at org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl.addDataInternal(DataStreamerImpl.java:646)
> at org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl.addDataInternal(DataStreamerImpl.java:631)
> at org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl.addData(DataStreamerImpl.java:753)
> at com.sc.extr.cache.PureIgniteDynamicRowStorage.putIfAbsent(PureIgniteDynamicRowStorage.java:83)
> at com.sc.extr.cache.PureIgniteDynamicRowStorage.addRowOnKey(PureIgniteDynamicRowStorage.java:160)
> at com.sc.extr.tree.MultiCacheTreeBuilder.addRootRowToCache(MultiCacheTreeBuilder.java:409)
> at com.sc.extr.tree.MultiCacheTreeBuilder.parentRev1to1(MultiCacheTreeBuilder.java:237)
> at com.sc.extr.tree.MultiCacheTreeBuilder.addRowToCache(MultiCacheTreeBuilder.java:333)
> at com.sc.extr.tree.MultiCacheTreeBuilder.parentRev(MultiCacheTreeBuilder.java:274)
> at com.sc.extr.tree.MultiCacheTreeBuilder.addRow(MultiCacheTreeBuilder.java:379)
> at com.sc.extr.tree.MultiCacheTreeBuilder.process(MultiCacheTreeBuilder.java:206)
> at com.sc.bi.workflow.WorkTransformer.processOne(WorkTransformer.java:84)
> at com.sc.bi.workflow.WorkTransformer.doWork(WorkTransformer.java:145)
> at com.sc.bi.workflow.WorkTransformer.processQueue(WorkTransformer.java:210)
> at com.sc.bi.workflow.WorkTransformer.run(WorkTransformer.java:169)
> Caused by: class org.apache.ignite.IgniteCheckedException: Data streamer has been cancelled: DataStreamerImpl [bufLdrSzPerThread=4096, rcvr=org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl$IsolatedUpdater@381b03ed, ioPlcRslvr=null, cacheName=PERSON.PTINTN, bufSize=512, parallelOps=0, timeout=-1, autoFlushFreq=0, bufMappings=ConcurrentHashMap {03e74462-12ec-4140-b9fb-a975572ac3bb=Buffer [node=TcpDiscoveryNode [id=03e74462-12ec-4140-b9fb-a975572ac3bb, consistentId=b01eb38b-7728-4e43-a697-0bc52f872e44, addrs=ArrayList [127.0.0.1, 172.27.179.112], sockAddrs=HashSet [SOFTBI-DEV.sc.com/172.27.179.112:47500, /127.0.0.1:47500], discPort=47500, order=1, intOrder=1, lastExchangeTime=1590614830815, loc=true, ver=2.8.1#20200521-sha1:86422096, isClient=false], isLocNode=true, idGen=0, sem=java.util.concurrent.Semaphore@2a869d9[Permits = 64], perNodeParallelOps=64, entriesCnt=2048, locFutsSize=0, reqsSize=0]}, cacheObjProc=GridProcessorAdapter [], cacheObjCtx=org.apache.ignite.internal.processors.cache.CacheObjectContext@2a5313b0, cancelled=true, cancellationReason=null, failCntr=0, activeFuts=GridConcurrentHashSet [GridFutureAdapter [ignoreInterrupts=false, state=INIT, res=null, hash=2102798044], GridFutureAdapter [ignoreInterrupts=false, state=INIT, res=null, hash=1195632760], GridFutureAdapter [ignoreInterrupts=false, state=INIT, res=null, hash=370791970], GridFutureAdapter [ignoreInterrupts=false, state=INIT, res=null, hash=420732031], GridFutureAdapter [ignoreInterrupts=false, state=INIT, res=null, hash=1453517070]], jobPda=null, depCls=null, fut=DataStreamerFuture [super=GridFutureAdapter [ignoreInterrupts=false, state=INIT, res=null, hash=1165180540]], publicFut=IgniteFuture [orig=DataStreamerFuture [super=GridFutureAdapter [ignoreInterrupts=false, state=INIT, res=null, hash=1165180540]]], disconnectErr=null, closed=true, lastFlushTime=1590629894701, skipStore=false, keepBinary=false, maxRemapCnt=32, remapSem=java.util.concurrent.Semaphore@6e6f060b[Permits = 2147483647], remapOwning=false]
> at org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl.closeEx(DataStreamerImpl.java:1347)
> at org.apache.ignite.internal.processors.datastreamer.DataStreamerImpl.closeEx(DataStreamerImpl.java:1318)
> at org.apache.ignite.internal.processors.datastreamer.DataStreamProcessor.onKernalStop(DataStreamProcessor.java:155)
> at org.apache.ignite.internal.IgniteKernal.stop0(IgniteKernal.java:2551)
> at org.apache.ignite.internal.IgniteKernal.stop(IgniteKernal.java:2499)
> at org.apache.ignite.internal.IgnitionEx$IgniteNamedInstance.stop0(IgnitionEx.java:2650)
> at org.apache.ignite.internal.IgnitionEx$IgniteNamedInstance.stop(IgnitionEx.java:2613)
> at org.apache.ignite.internal.IgnitionEx.stop(IgnitionEx.java:339)
> at org.apache.ignite.failure.StopNodeFailureHandler$1.run(StopNodeFailureHandler.java:36)
> at java.lang.Thread.run(Thread.java:748)
--
This message was sent by Atlassian Jira
(v8.3.4#803005)