Details
Blocker Description
I have a only one DIST_SYNC cache, most of the JVM in the cluster are configured with capacityFactor = 0 (like the distibutedlocalstorage=false property of Coherence) and some node are configured with capacityFactor>0 (for instance 1000). We are talking about 100 nodes with capacityFactor=0 and 4 nodes of the other kind, al the cluster is indide one single "site/rack". Partition Handling is off, numOwners is 1.
When all the nodes with capacityFactor > 0 are down the cluster comes to a degraded state
the ploblem is that even if nodes with capacityFactor>0 are up again the cluster does not recover, a full restart is needed
If I enable partition-handling AvailablyExceptions start to be throw and I think is the expected behaviour (see the "Infinispan User Guide").
I think this is the problem and it is a bug:
14/11/17 09:27:25 WARN topology.CacheTopologyControlCommand: ISPN000071: Caught exception when handling command CacheTopologyControlCommand{cache=shared, type=JOIN, sender=testserver1@xxxxxxx-22311, site-id=xxx, rack-id=xxx, machine-id=24 bytes, joinInfo=CacheJoinInfo{consistentHashFactory=org.infinispan.distribution.ch.impl.TopologyAwareConsistentHashFactory@78b791ef, hashFunction=MurmurHash3, numSegments=60, numOwners=1, timeout=120000, totalOrder=false, distributed=true}, topologyId=0, rebalanceId=0, currentCH=null, pendingCH=null, availabilityMode=null, throwable=null, viewId=3}
java.lang.IllegalArgumentException: A cache topology's pending consistent hash must contain all the current consistent hash's members
at org.infinispan.topology.CacheTopology.<init>(CacheTopology.java:48)
at org.infinispan.topology.CacheTopology.<init>(CacheTopology.java:43)
at org.infinispan.topology.ClusterCacheStatus.startQueuedRebalance(ClusterCacheStatus.java:631)
at org.infinispan.topology.ClusterCacheStatus.queueRebalance(ClusterCacheStatus.java:85)
at org.infinispan.partionhandling.impl.PreferAvailabilityStrategy.onJoin(PreferAvailabilityStrategy.java:22)
at org.infinispan.topology.ClusterCacheStatus.doJoin(ClusterCacheStatus.java:540)
at org.infinispan.topology.ClusterTopologyManagerImpl.handleJoin(ClusterTopologyManagerImpl.java:123)
at org.infinispan.topology.CacheTopologyControlCommand.doPerform(CacheTopologyControlCommand.java:158)
at org.infinispan.topology.CacheTopologyControlCommand.perform(CacheTopologyControlCommand.java:140)
at org.infinispan.remoting.transport.jgroups.CommandAwareRpcDispatcher$4.run(CommandAwareRpcDispatcher.java:278)
at java.util.concurrent.ThreadPoolExecutor.runWorker(ThreadPoolExecutor.java:1142)
at java.util.concurrent.ThreadPoolExecutor$Worker.run(ThreadPoolExecutor.java:617)
at java.lang.Thread.run(Thread.java:745)
After that error every "put" results in:
14/11/17 09:27:27 ERROR interceptors.InvocationContextInterceptor: ISPN000136: Execution error
org.infinispan.util.concurrent.TimeoutException: Timed out waiting for topology 1
at org.infinispan.statetransfer.StateTransferLockImpl.waitForTransactionData(StateTransferLockImpl.java:93)
at org.infinispan.interceptors.base.BaseStateTransferInterceptor.waitForTransactionData(BaseStateTransferInterceptor.java:96)
at org.infinispan.statetransfer.StateTransferInterceptor.handleNonTxWriteCommand(StateTransferInterceptor.java:188)
at org.infinispan.statetransfer.StateTransferInterceptor.visitPutKeyValueCommand(StateTransferInterceptor.java:95)
at org.infinispan.commands.write.PutKeyValueCommand.acceptVisitor(PutKeyValueCommand.java:71)
at org.infinispan.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:98)
at org.infinispan.interceptors.CacheMgmtInterceptor.updateStoreStatistics(CacheMgmtInterceptor.java:148)
at org.infinispan.interceptors.CacheMgmtInterceptor.visitPutKeyValueCommand(CacheMgmtInterceptor.java:134)
at org.infinispan.commands.write.PutKeyValueCommand.acceptVisitor(PutKeyValueCommand.java:71)
at org.infinispan.interceptors.base.CommandInterceptor.invokeNextInterceptor(CommandInterceptor.java:98)
at org.infinispan.interceptors.InvocationContextInterceptor.handleAll(InvocationContextInterceptor.java:102)
at org.infinispan.interceptors.InvocationContextInterceptor.handleDefault(InvocationContextInterceptor.java:71)
at org.infinispan.commands.AbstractVisitor.visitPutKeyValueCommand(AbstractVisitor.java:35)
at org.infinispan.commands.write.PutKeyValueCommand.acceptVisitor(PutKeyValueCommand.java:71)
at org.infinispan.interceptors.InterceptorChain.invoke(InterceptorChain.java:333)
at org.infinispan.cache.impl.CacheImpl.executeCommandAndCommitIfNeeded(CacheImpl.java:1576)
at org.infinispan.cache.impl.CacheImpl.putInternal(CacheImpl.java:1054)
at org.infinispan.cache.impl.CacheImpl.put(CacheImpl.java:1046)
at org.infinispan.cache.impl.CacheImpl.put(CacheImpl.java:1646)
at org.infinispan.cache.impl.CacheImpl.put(CacheImpl.java:245)
This is the actual configuration:
GlobalConfiguration globalConfig = new GlobalConfigurationBuilder() .globalJmxStatistics() .allowDuplicateDomains(true) .cacheManagerName(instanceName) .transport() .defaultTransport() .clusterName(clustername) .addProperty("configurationFile", configurationFile) (udp for my cluster, approx 100 machines) .machineId(instanceName) .siteId("site1") .rackId("rack1") .nodeName(serviceName + "@" + instanceName) .remoteCommandThreadPool().threadPoolFactory(CachedThreadPoolExecutorFactory.create()) .build(); Configuration wildcard = new ConfigurationBuilder() .locking().lockAcquisitionTimeout(lockAcquisitionTimeout) .concurrencyLevel(10000).isolationLevel(IsolationLevel.READ_COMMITTED).useLockStriping(true) .clustering() .cacheMode(CacheMode.DIST_SYNC) .l1().lifespan(l1ttl) .hash().numOwners(numOwners).capacityFactor(capacityFactor) .partitionHandling().enabled(false) .stateTransfer().awaitInitialTransfer(false).timeout(initialTransferTimeout).fetchInMemoryState(false) .storeAsBinary().enabled(true).storeKeysAsBinary(false).storeValuesAsBinary(true) .jmxStatistics().enable() .unsafe().unreliableReturnValues(true) .build();
One workaround is to set capacityFactor = 1 instead of 0, but I do not want "simple-nodes" (with less RAM) to becaome key-owners
For me this is a showstopper problem
Attachments
Issue Links
- is duplicated by
-
ISPN-11227 Cluster fails to startup due to initial state transfer timing out
-
- Closed
-
