I'm trying to use session replication but without lock contention.
So I set up the infinispan "web" cache-container like this:
<cache-container name="web" default-cache="repl" module="org.wildfly.clustering.web.infinispan">
<transport lock-timeout="60000"/>
<replicated-cache name="repl">
<locking isolation="READ_COMMITTED"/>
<transaction locking="OPTIMISTIC" mode="NONE"/>
<file-store/>
</replicated-cache>
</cache-container>
But the session is not replicating across the cluster.
It replicates only if I use mode="BATCH" and default transaction locking PESSIMISTIC. But with this strategy does not perform well with long requests (about 2~3 seconds) and a lot of concurrent access (one request locks others requests because the first request owns the session lock).
Is there any way to replicate the session across the cluster without using transaction and consequently without session lock?
Thanks!
Related
I am using the 11.0.7 Infinispan cache store which is configured with an XML file. I have imported that cache module in a Wildlfy server. I am using this dependency in my application. But when I try to fetch cache elements from cache, it gives me the exception ClassNotFound.
I have used this configuration:
<local-cache name="TaskStoreCache" statistics="false">
<locking acquire-timeout="60000" />
<persistence passivation="false">
<rocksdb-store path=" C:\CacheStore\Data\TaskStoreCache" preload="false" shared="false"
purge="false" read-only="false">
<expiration path="C: \CacheStore\Expired\TaskStoreCache"/>
</rocksdb-store>
</persistence>
<memory max-count="500"/>
<encoding media-type="application/x-java-object"/>
</local-cache>
And I used this serialization:
<serialization marshaller="org.infinispan.commons.marshall.JavaSerializationMarshaller">
<white-list>
<regex>com.xyz.cache.*</regex>
<regex>java.util.*</regex>
<regex>java.lang.*</regex>
</white-list>
</serialization>
I am sure it is a class loading related issue. Please help me with that.
You should double-check how the dependency is wrapped into your application (war/ear), and make sure the scope is correct.
If you did not pack it within your application, the module has to be present on your wildfly and known to your application (e.g. jboss-deployment-structure.xml)
I am attempting to use the jdbc-store type for my session cache in Wildfly 15+.
I ran the following commands to configure my standalone-full-ha.xml configuration file:
/subsystem=infinispan/cache-container=web/invalidation-cache=jdbc/:add(mode=SYNC)
/subsystem=infinispan/cache-container=web/invalidation-cache=jdbc/store=none:remove(){allow-resource-service-restart=true}
/subsystem=infinispan/cache-container=web/invalidation-cache=jdbc/store=jdbc/:add(data-source="...",passivation=false,shared=true){allow-resource-service-restart=true}
/subsystem=infinispan/cache-container=web/invalidation-cache=jdbc/component=transaction/:add()
/subsystem=infinispan/cache-container=web/invalidation-cache=jdbc/component=transaction/:write-attribute(name=mode,value=BATCH)
/subsystem=infinispan/cache-container=web:write-attribute(name=default-cache,value=jdbc)
... which produces the following in the configuration file:
<cache-container name="web" default-cache="jdbc" module="org.wildfly.clustering.web.infinispan">
<transport lock-timeout="60000"/>
<invalidation-cache name="jdbc">
<transaction mode="BATCH"/>
<jdbc-store data-source="..." passivation="false" shared="true">
<table/>
</jdbc-store>
</invalidation-cache>
<distributed-cache name="dist">
<locking isolation="REPEATABLE_READ"/>
<transaction mode="BATCH"/>
<file-store/>
</distributed-cache>
</cache-container>
It looks like I've configured the cache correctly using JBoss CLI, but when the cluster instances start-up, the session store table is not created in the database, even though everything starts up properly.
My question is, is there something that I should be setting in the <table/> element that I'm just overlooking? Looking at the documentation, I don't see any required attributes, or anything about auto-creation.
I've looked at previous examples of how to achieve this in Wildfly 11, but the string-keyed-jdbc-store element no longer seems to be valid. I know the Infinispan documentation mentions the create-on-start attribute on the string-keyed-table element, but this configuration is so wildly different in Wildfly that it's completely unhelpful.
Add
<property name="createTableOnStart">
true
</property>
in your jdbc-store element
and make sure your web.xml is <distributable />
Is it possible to configure failover URI for native Artemis server in Wildfly 14? I'd like to create pooled connection factory with url like (tcp://localhost:61616,tcp://localhost:61617). As far as I know wildfly creates connection factory from connector host and port. I use Wildfly 14.0.1.Final and Artemis 2.6.3.
Update
<remote-connector name="remote-artemis-master" socket-binding="remote-artemis-master" />
<remote-connector name="remote-artemis-slave" socket-binding="remote-artemis-slave" />
<pooled-connection-factory
ha="true"
name="activemq-ra"
connectors="remote-artemis-master remote-artemis-slave"
entries="java:/JmsXA java:jboss/DefaultJMSConnectionFactory"
transaction="xa"
user="user"
password="password" />
Update
When master node stops the slave becomes live but the jee app is unable to send/consume messages 30 for seconds. After this period everything works fine.
The syntax (tcp://localhost:61616,tcp://localhost:61617) is just a way to configure multiple initial connectors via a URL. You can accomplish the same thing in Wildfly by defining multiple remote-connector elements and referencing those in the connectors attribute of the pooled-connection-factory.
I'm migrating from Wildfly 8.2 to 10.1 Unfortunately, I'm encountering problems with Infinispan TreeCache.
Here are several issues:
Invocation batching is no longer supported in Wildfly 10
configuration
Here's my config:
<subsystem xmlns="urn:jboss:domain:infinispan:4.0">
...
<cache-container name="my_container" default-cache="my_tree_cache" jndi-name="java:jboss/my_container">
<transport lock-timeout="60000"/>
<local-cache name="my_cache"/>
<local-cache name="my_tree_cache" batching="true"/>
</cache-container>
</subsystem>
Error on startup:
> Caused by: javax.xml.stream.XMLStreamException: ParseError at [row,col]:[345,17]
> Message: WFLYCTL0197: Unexpected attribute 'batching' encountered
If I remove "batching" attribute. I get this error:
com.daiwacm.modjsf.dataaccess.DataException: getTreeCache has failed for jndi value (my_tree_cache)
Caused by: org.infinispan.commons.CacheConfigurationException: invocationBatching is not
enabled for cache 'my_tree_cache'. Make sure this is enabled by calling configurationBuilder.invocationBatching().enable()
If I set batching programmatically:
Context context = new InitialContext();
CacheContainer cacheContainer = (CacheContainer) context.lookup(jndiName);
TreeCacheFactory tcf = new TreeCacheFactory();
Cache cache = cacheContainer.getCache(cacheName);
cache.getCacheManager().defineConfiguration(cacheName,
new ConfigurationBuilder().read(cache.getCacheConfiguration()).invocationBatching().enable().build());
TreeCache treeCache = tcf.createTreeCache(cache);
I get this error:
> Caused by: org.infinispan.commons.CacheConfigurationException:
> ISPN000381: This configuration is not supported for simple cache
> at org.infinispan.configuration.cache.ConfigurationBuilder.validateSimpleCacheConfiguration(ConfigurationBuilder.java:219)
> ...
> at org.infinispan.configuration.cache.InvocationBatchingConfigurationBuilder.build(InvocationBatchingConfigurationBuilder.java:12)
> ...
Don't set the configuration programmatically; I am not sure this is a valid approach, despite it seems to ~work.
The configuration option you're looking for is
<local-cache name="my_cache">
<transaction transaction-mode="BATCH" />
</local-cache>
(please consult the schema in docs/schema/jboss-as-infinispan_4_0.xsd should you have any doubts)
The last problem is that for local caches, WF automatically enables certain optimizations when it's possible. When you redefine the cache programmatically, this optimization (simple cache) is set on. So you'd have to set
new ConfigurationBuilder().read(cache.getCacheConfiguration())
.simpleCache(false)
.invocationBatching().enable()
Hibernate search default infinispan configuration store indexes in memory,you have to reindex everything once you shutdown application.
I read infinispan document, there is a way to store index into a infinispan file store. After I do googling around and around, I still don't know how to configure it.
You can check Infinispan user guide chapters 5 (Persistence) and 16 (Infinispan as a storage for Lucene indexes). Chapter numbers are from Infinispan 8.2. Hibernate search also provides a "default-hibernatesearch-infinispan.xml" file to start with. You basically need to add persistence to metadata and actual index caches. Here is the one I use for index cache:
<distributed-cache name="LuceneIndexesData" mode="SYNC" remote-timeout="25000">
<transaction mode="NONE"/>
<state-transfer enabled="true" timeout="480000" await-initial-transfer="true"/>
<indexing index="NONE"/>
<locking striping="false" acquire-timeout="10000" concurrency-level="500" write-skew="false"/>
<eviction max-entries="-1" strategy="NONE"/>
<expiration max-idle="-1"/>
<persistence passivation="false">
<jdbc:string-keyed-jdbc-store preload="true" fetch-state="true" read-only="false" purge="false">
<jdbc:data-source jndi-url="java:comp/env/jdbc/..."/>
<jdbc:string-keyed-table drop-on-exit="false" create-on-start="true" prefix="ISPN_STRING_TABLE">
<jdbc:id-column name="ID" type="VARCHAR(255)"/>
<jdbc:data-column name="DATA" type="MEDIUMBLOB"/>
<jdbc:timestamp-column name="TIMESTAMP" type="BIGINT"/>
</jdbc:string-keyed-table>
<property name="key2StringMapper">org.infinispan.lucene.LuceneKey2StringMapper</property>
<write-behind/>
</jdbc:string-keyed-jdbc-store>
</persistence>
</distributed-cache>
This example uses JDBC because it works on a dynamic cluster. You need to replace "jdbc:string-keyed=jdbc-store" with a "file-store" if you want to store the index a file.