This document describes how to virtualize Java applications in Java by hosting multiple "pseudo JVMs" within a single JVM. This allows deploying distributed applications for testing purposes. Key points covered include:
- Using custom classloaders and system properties to isolate "pseudo JVMs" and simulate distributed environments.
- Frameworks like GridKit that enable starting whole application topologies within JUnit tests for behaviors testing.
- Techniques for testing features like serialization, data routing, and cross-version compatibility.
- Later extensions to deploy virtual nodes across real servers using SSH for performance and deployment testing of distributed systems.
2. Virtualizing Java in Java
Single JVM hosting
•
•
•
•
•
multiple “pseudo JVMs”, which
Have independent system properties
Have independent statics
May have different classpath
May be forcefully terminated or
suspended/resumed
You can deploy whole distributed application
topology inside of JUnit test
3. All started with Oracle Coherence
Native Java distributed peer-to-peer key/value storage
With a lot of extension points
Distributed data processing
Listeners for remote events
Read-through / write-through patterns
Pluggable serialization
And inability to start two cluster members in one JVM
4. And we need to test
Behavior of very specific features
Serialization/deserialization defects
Data routing and collocation aspects
Code meant for distributed execution
Threading aspects
Classpath differences between cluster
processes
Cluster configuration tweaks
5. Multiplexing singletons
Custom classloader
Force to reload classes already loaded by
parent (second copy of class loaded)
Got our first cluster-in-JVM alive
InheritentThreadLocal to add fancy stuff
– Multiplexing system properties
– Multiplexing multiplex console output
6. Heated competition
Three open source “cluster virtualizing” effort around
Oracle Coherence
GridKit (ChTest)
code.google.com/p/gridkit
Oracle tools
github.com/coherence-community/oracle-tools
Little
www.littlegrid.net
7. “Distributing” test case
How to start The Application on “virtual cluster”?
Old school
Main classes and command line arguments
But if you need to do verification inside of
vinode?
A separate main for each test case?
8. But how all this is relevant to me?
Normal client/server application
You can use your real main classes instead of mocking server
Hadoop / HBase / Cassandra
Distributed
Deployment unfriendly
Ship with single node – cut down versions
9. “Distributing” test case
Transparent remotting
client1.exec(new Runnable() {
@Override
public void run() {
NamedCache cache = CacheFactory.getCache(cacheName);
Assert.assertNull(cache.get("A"));
cache.put("A", "aaa");
}
});
client2.exec(new Runnable() {
@Override
public void run() {
NamedCache cache = CacheFactory.getCache(cacheName);
Assert.assertEquals("aaa", cache.get("A"));
}
});
10. We want to test more!
We are in control class loading
— let’s tweak classpath on flight
Inject resources
Remove server classes from client
Test different codebase versions
12. Cross version tests
Master JVM, client regression test pack
[trunk version]
Case
Client
[version X]
Client
[version Y]
Server
[version Z]
13. Managing artifacts
… a bunch of black magic to find local repo
and managing classpath as easy as …
<plugin>
<groupId>org.apache.maven.plugins</groupId>
<artifactId>maven-dependency-plugin</artifactId>
<version>2.8</version>
<executions>
<execution>
<id>viconcurrent-0.7.15</id>
<phase>test-compile</phase>
<goals>
<goal>get</goal>
</goals>
<configuration>
<artifact>org.gridkit.lab:viconcurrent:0.7.15</artifact>
</configuration>
</execution>
</executions>
</plugin>
14. Managing artifacts
How to get needed artifact on local disk
- Maven will disallow two versions of same artifact
- but we can trick it …
ViNode node;
…
node.x(MAVEN).replace("org.gridkit.lab", "viconcurrent", "0.7.15");
Transitive dependencies are not included, though.
15. Transparent remotting
Own implementation of RMI
Standard Java serialization
Serializing of anonymous Runnable/Callable
Auto export of Remote interfaces
Writing distributed code by convention!
16. Bidirectional communications
public interface RemotePut extends Remote {
public void put(Object key, Object value);
}
public interface RemotePut extends Remote {
@SuppressWarnings("unused")
@Test
public void bidirectional_remoting() {
// Present for typical single node cluster
cloud.all().presetFastLocalCluster();
public void put(Object key, Object value);
cloud.node("storage.**").localStorage(true);
cloud.node("client.**").localStorage(false);
}
// Simulates DefaultCacheServer based process
cloud.node("storage.**").autoStartServices();
// declaring specific nodes to be created
CohNode storage = cloud.node("storage.1");
CohNode client1 = cloud.node("client.1");
CohNode client2 = cloud.node("client.2");
RemotePut remoteService =
client1.exec(new Callable<RemotePut>() {
@Override
public RemotePut call() {
final NamedCache cache = CacheFactory.getCache(cacheName);
// now we have 3 specific nodes in cloud
// all of then will be initialized in parallel
cloud.all().ensureCluster();
final String cacheName = "distr-a";
return new RemotePut() {
@Override
public void put(Object key, Object value) {
cache.put(key, value);
}
};
RemotePut remoteService =
client1.exec(new Callable<RemotePut>() {
@Override
public RemotePut call() {
final NamedCache cache = CacheFactory.getCache(cacheName);
return new RemotePut() {
@Override
public void put(Object key, Object value) {
cache.put(key, value);
}
};
}
});
}
});
remoteService.put("A", "aaa");
remoteService.put("A", "aaa");
client2.exec(new Runnable() {
@Override
public void run() {
NamedCache cache = CacheFactory.getCache(cacheName);
Assert.assertEquals("aaa", cache.get("A"));
}
});
}
17. Bidirectional communications
Extending java.rmi.Remotexec
will mark interface for auto export
public interface RemotePut extends Remote {
public void put(Object key, Object value);
}
RemotePut remoteService =
client1.exec(new Callable<RemotePut>() {
@Override
public RemotePut call() {
final NamedCache cache = CacheFactory.getCache(cacheName);
return new RemotePut() {
@Override
public void put(Object key, Object value) {
cache.put(key, value);
}
};
}
});
remoteService.put("A", "aaa");
Here we got a remote stub, not a
real implementation of interface
Unlike Java RMI, there is no need to
declare RemoteException for
every method
Result of callable will be serialized
and transferred to caller
Objects implementing remote
interfaces are automatically replaced
with remote stub during serialization
Call to a stub, will be converted to
“remote” call to instance we have
created in “virtualized” node few
lines above
18. Sneak peek: Instrumentation
System.exit() – is still fatal
Some cases need “virtual time”
Tweaking monolithic code
Fault injection
Mock injection
23. Virtual stuff is so good
Managing virtual nodes in deterministic way, in
Java, having all luxury of exception handling and
richness of libraries – feeling were so good …
I wished, I could rollout JVMs across real servers
24. Your network is Big JVM
Same API
3 types of nodes: in-process, local, remote
Transparent remotting
SSH to manage remote server
Automatic classpath replication (with caching)
Zero infrastructure
Any OS for master host
SSHd + JVM for slave hosts
25. New opportunities
Performance testing
deploy system under test
deploy load generators
deploy monitoring agents
gather all result in one place
Deployment (remote execution task for ANT)
Replace your putty with Java IDE
log scrapping
parallel execution
26. As easy as …
@Test
public void remote_hello_world() throws InterruptedException
{
ViManager cloud = CloudFactory.createSimpleSshCloud();
cloud.node("myserver.uk.db.com");
cloud.node("**").exec(new Callable<Void>() {
@Override
public Void call() throws Exception
{
String localHost = InetAddress.getLocalHost().toString();
System.out.println("Hi! I'm running on " + localHost);
return null;
}
});
}
27. Behind the scene
•
•
•
•
JSCh – SSH client (slightly patched)
Collected classpath artifacts
SCP jars to remote target
Start remote agent (JVM)
stdOut / stdIn for master – agent communications
• Agent start slave process
• Slave start RMI node and connects (TCP) to master
agent acts as TCP proxy
28. Death clock is ticking
Master JVM kills slave processes, unless
SSH session was interrupted
someone kill -9 master JVM
master JVM has crashed (e.g. under debuger)
Death clock is ticking on slave though
if master is not responding
slave process will terminate itself
29. Performance testing
•
•
•
•
•
Master JVM is running on CI
JUnit to start test
Loaded generator farm is deployed on test servers
Monitoring agents deployed on application cluster
Metrics are buffered locally, then send to master and processed
Real test
• Four servers – application
• 50 servers – load farm
• Over 200 – slaves processes
30. Coding for 200+ processes
Driver - concept
• Driver – Java interface encapsulates test
action
• One way methods
• Friendly for remotting for parallel invokation
+ some utility for parallel execution, workflow
etc
Example:
https://gridkit.googlecode.com/svn/grid-lab/trunk/examples/zk-benchmark-sample
33. Thank you
http://blog.ragozin.info
- my articles
http://code.google.com/p/gridkit
http://github.com/gridkit
- my open source code
http://aragozin.timepad.ru
- tech meetups in Moscow
Alexey Ragozin
alexey.ragozin@gmail.com