FD, replication timeouts, and shunning
quinine Jul 10, 2007 6:02 PMThe short version of this question is: I have two nodes, A & B, with TreeCache hibernate 2nd-level cache replication. Node B will have several GCs during the day that make it unresponsive for minutes at a time. During these, Node A throws ReplicationExceptions. I want to set my jbosscache/jgroups configuration to enable node A to withstand these GC events on node B, either by temporarily shunning or by having appropriate timeouts, or some combination of both.
Long Version:
I have inherited a jboss cluster. I have been doing a great deal of reading, I have gone through the jgroups, jbosscache, & hibernate jbosscache wikis here on jboss.org.
We have 2 nodes , A & B, in a cluster. Both are running:
JBoss 4.0.5
JBossCache 1.4.1.SP3
JGroups 2.4.1-SP1.
The jboss-config.xml for the hibernate treecache (identical on both):
<?xml version="1.0" encoding="UTF-8"?> <server> <mbean code="org.jboss.cache.TreeCache" name="jboss.cache:service=HibernateTreeCache"> <depends>jboss:service=Naming</depends> <depends>jboss:service=TransactionManager</depends> <attribute name="ClusterName">Hibernate-${jboss.partition.name:Cluster}</attribute> <attribute name="IsolationLevel">READ_COMMITTED</attribute> <attribute name="CacheMode">REPL_SYNC</attribute> <attribute name="UseRegionBasedMarshalling">false</attribute> <attribute name="InactiveOnStartup">false</attribute> <attribute name="TransactionManagerLookupClass">org.jboss.cache.BatchModeTransactionManagerLookup</attribute> <attribute name="ClusterConfig"> <config> <TCP bind_addr="${partition.tcphost:HIB3-MISCONFIGURED}" start_port="${partition.tcpport.hib3:HIB3-MISCONFIGURED}" loopback="false" tcp_nodelay="false" up_thread="false" down_thread="false"/> <TCPPING initial_hosts="${partition.tcphosts.hib3:HIB3-MISCONFIGURED}" port_range="3" timeout="3500" num_initial_members="3" up_thread="false" down_thread="false"/> <MERGE2 min_interval="20000" max_interval="100000" down_thread="false" up_thread="false"/> <FD_SOCK down_thread="false" up_thread="false"/> <FD shun="true" down_thread="false" up_thread="false" timeout="20000" max_tries="5"/> <VERIFY_SUSPECT timeout="1500" down_thread="false" up_thread="false"/> <pbcast.NAKACK up_thread="false" down_thread="false" gc_lag="100" retransmit_timeout="60000"/> <pbcast.STABLE desired_avg_gossip="50000" up_thread="false" down_thread="false" /> <pbcast.GMS join_timeout="5000" join_retry_timeout="2000" shun="true" print_local_addr="true" down_thread="false" up_thread="false"/> <pbcast.STATE_TRANSFER up_thread="false" down_thread="false"/> </config> </attribute> <attribute name="FetchInMemoryState">false</attribute> <attribute name="InitialStateRetrievalTimeout">20000</attribute> <attribute name="SyncReplTimeout">20000</attribute> <attribute name="LockAcquisitionTimeout">15000</attribute> <attribute name="BuddyReplicationConfig"> <config> <buddyReplicationEnabled>false</buddyReplicationEnabled> <buddyLocatorClass>org.jboss.cache.buddyreplication.NextMemberBuddyLocator</buddyLocatorClass> <buddyLocatorProperties> numBuddies = 1 ignoreColocatedBuddies = true </buddyLocatorProperties> <buddyPoolName>default</buddyPoolName> <buddyCommunicationTimeout>2000</buddyCommunicationTimeout> <autoDataGravitation>false</autoDataGravitation> <dataGravitationRemoveOnFind>true</dataGravitationRemoveOnFind> <dataGravitationSearchBackupTrees>true</dataGravitationSearchBackupTrees> </config> </attribute> </mbean> </server>
Our cluster-service.xml:
<?xml version="1.0" encoding="UTF-8"?> <server> <mbean code="org.jboss.ha.framework.server.ClusterPartition" name="jboss:service=${jboss.partition.name:DefaultPartition}"> <attribute name="PartitionName">${jboss.partition.name:DefaultPartition}</attribute> <attribute name="NodeAddress">${jboss.bind.address}</attribute> <attribute name="DeadlockDetection">False</attribute> <attribute name="StateTransferTimeout">30000</attribute> <attribute name="PartitionConfig"> <Config> <TCP bind_addr="${partition.tcphost:CLUSTERCONFIG-MISCONFIGURED}" start_port="${partition.tcpport.cluster:CLUSTERCONFIG-MISCONFIGURED}" loopback="false" recv_buf_size="2000000" send_buf_size="640000" tcp_nodelay="true" up_thread="true" down_thread="true"/> <TCPPING initial_hosts="${partition.tcphosts.cluster:CLUSTERCONFIG-MISCONFIGURED}" port_range="3" timeout="3500" num_initial_members="3" up_thread="true" down_thread="true"/> <MERGE2 min_interval="10000" max_interval="20000" /> <FD_SOCK down_thread="true" up_thread="true"/> <FD shun="true" up_thread="true" down_thread="true" timeout="10000" max_tries="5"/> <VERIFY_SUSPECT timeout="3000" down_thread="true" up_thread="true" /> <pbcast.NAKACK up_thread="true" down_thread="true" gc_lag="100" retransmit_timeout="300,600,1200,2400,4800"/> <pbcast.STABLE desired_avg_gossip="20000" max_bytes="400000" down_thread="true" up_thread="true" /> <pbcast.GMS join_timeout="5000" join_retry_timeout="2000" shun="true" print_local_addr="true" up_thread="true" down_thread="true"/> <FC max_credits="2000000" down_thread="true" up_thread="true" min_threshold="0.10"/> <FRAG2 frag_size="60000" down_thread="true" up_thread="true"/> <pbcast.STATE_TRANSFER up_thread="true" down_thread="true"/> </Config> </attribute> <depends>jboss:service=Naming</depends> </mbean> <mbean code="org.jboss.ha.hasessionstate.server.HASessionStateService" name="jboss:service=HASessionState"> <depends>jboss:service=Naming</depends> <!-- We now inject the partition into the HAJNDI service instead of requiring that the partition name be passed --> <depends optional-attribute-name="ClusterPartition" proxy-type="attribute">jboss:service=${jboss.partition.name:DefaultPartition}</depends> <!-- JNDI name under which the service is bound --> <attribute name="JndiName">/HASessionState/Default</attribute> <!-- Max delay before cleaning unreclaimed state. Defaults to 30*60*1000 => 30 minutes --> <attribute name="BeanCleaningDelay">0</attribute> </mbean> <mbean code="org.jboss.ha.jndi.HANamingService" name="jboss:service=HAJNDI"> <!-- We now inject the partition into the HAJNDI service instead of requiring that the partition name be passed --> <depends optional-attribute-name="ClusterPartition" proxy-type="attribute">jboss:service=${jboss.partition.name:DefaultPartition}</depends> <!-- Bind address of bootstrap and HA-JNDI RMI endpoints --> <attribute name="BindAddress">${jboss.bind.address}</attribute> <!-- Port on which the HA-JNDI stub is made available --> <attribute name="Port">1100</attribute> <!-- RmiPort to be used by the HA-JNDI service once bound. 0 => auto. --> <attribute name="RmiPort">1101</attribute> <!-- Accept backlog of the bootstrap socket --> <attribute name="Backlog">50</attribute> <!-- The thread pool service used to control the bootstrap and auto discovery lookups --> <depends optional-attribute-name="LookupPool" proxy-type="attribute">jboss.system:service=ThreadPool</depends> <!-- A flag to disable the auto discovery via multicast --> <attribute name="DiscoveryDisabled">false</attribute> <!-- Set the auto-discovery bootstrap multicast bind address. If not specified and a BindAddress is specified, the BindAddress will be used. --> <attribute name="AutoDiscoveryBindAddress">${jboss.bind.address}</attribute> <!-- Multicast Address and group port used for auto-discovery --> <attribute name="AutoDiscoveryAddress">${jboss.partition.udpGroup:230.0.0.4}</attribute> <attribute name="AutoDiscoveryGroup">1102</attribute> <!-- The TTL (time-to-live) for autodiscovery IP multicast packets --> <attribute name="AutoDiscoveryTTL">16</attribute> <!-- The load balancing policy for HA-JNDI --> <attribute name="LoadBalancePolicy">org.jboss.ha.framework.interfaces.RoundRobin</attribute> <!-- Client socket factory to be used for client-server RMI invocations during JNDI queries <attribute name="ClientSocketFactory">custom</attribute> --> <!-- Server socket factory to be used for client-server RMI invocations during JNDI queries <attribute name="ServerSocketFactory">custom</attribute> --> </mbean> <mbean code="org.jboss.invocation.jrmp.server.JRMPInvokerHA" name="jboss:service=invoker,type=jrmpha"> <attribute name="ServerAddress">${jboss.bind.address}</attribute> <attribute name="RMIObjectPort">4447</attribute> <!-- <attribute name="RMIClientSocketFactory">custom</attribute> <attribute name="RMIServerSocketFactory">custom</attribute> --> <depends>jboss:service=Naming</depends> </mbean> <!-- the JRMPInvokerHA creates a thread per request. This implementation uses a pool of threads --> <mbean code="org.jboss.invocation.pooled.server.PooledInvokerHA" name="jboss:service=invoker,type=pooledha"> <attribute name="NumAcceptThreads">1</attribute> <attribute name="MaxPoolSize">300</attribute> <attribute name="ClientMaxPoolSize">300</attribute> <attribute name="SocketTimeout">60000</attribute> <attribute name="ServerBindAddress">${jboss.bind.address}</attribute> <attribute name="ServerBindPort">4446</attribute> <attribute name="ClientConnectAddress">${jboss.bind.address}</attribute> <attribute name="ClientConnectPort">0</attribute> <attribute name="EnableTcpNoDelay">false</attribute> <depends optional-attribute-name="TransactionManagerService">jboss:service=TransactionManager</depends> <depends>jboss:service=Naming</depends> </mbean> <mbean code="org.jboss.cache.invalidation.bridges.JGCacheInvalidationBridge" name="jboss.cache:service=InvalidationBridge,type=JavaGroups"> <!-- We now inject the partition into the HAJNDI service instead of requiring that the partition name be passed --> <depends optional-attribute-name="ClusterPartition" proxy-type="attribute">jboss:service=${jboss.partition.name:DefaultPartition}</depends> <depends>jboss.cache:service=InvalidationManager</depends> <attribute name="InvalidationManager">jboss.cache:service=InvalidationManager</attribute> <attribute name="BridgeName">DefaultJGBridge</attribute> </mbean> </server>
We occasionally get ReplicationExceptions on A, and we have been able to verify that these occur during long (up to 4-5 minutes) GCs on B, where the jvm becomes unresponsive.
As I read the config snippets above, node A will not shun until it doesn't receive a heartbeat for at least (20x5 + 1.5) = 101.5 seconds, but the replication timeout is 20 seconds.
So my questions are -
1) If I change the config so A will shun B before SyncReplTimeout, will this prevent replication during the time where B is unresponsive (hence preventing the ReplicationExceptions)?
2) I then am expecting B to be shunned somewhate regularly, but I always want B to be able to rejoin the cluster when it becomes responsive again. From what I'm reading, this means setting shun=false. Without shunning, how do I prevent replication to the unresponsive node B?
3) Are there further caveats that I need to consider? Will I need to make similar timeout/config changes to the cluster-config.xml?
Thank you very much for your time.