Skip to content

Latest commit

 

History

History
539 lines (387 loc) · 21.4 KB

README.md

File metadata and controls

539 lines (387 loc) · 21.4 KB
Raw

Java API User's Guide

This document describes how to use the Simulacron Java API to provision and interact with simulated clusters.

For non-java development, it is recommended to use the standalone jar.

Getting Simulacron

Simulacron can be added to your application by using the following maven dependency:

<dependency>
  <groupId>com.datastax.oss.simulacron</groupId>
  <artifactId>simulacron-native-server</artifactId>
  <version>0.8.9</version>
</dependency>

The native-server module provides all of the functionality needed to interact with simulacron, but if you are also using the java driver you should consider depending on the driver-3x module which provides convenience mechanisms (such as avoiding class name clashing) for working with the driver. To install driver-3x:

<dependency>
  <groupId>com.datastax.oss.simulacron</groupId>
  <artifactId>simulacron-driver-3x</artifactId>
  <version>0.8.9</version>
</dependency>

Creating a Server

Server provides a point of entry into provisioning and de-provisioning simulated clusters.

To set up a server, simply do the following:

import com.datastax.oss.simulacron.server.Server;

class Test {
  static Server server = Server.builder().build();
}

Constructing a Server will register a netty EventLoopGroup and Timer under the covers. Alternatively you may provide your own EventLoopGroup as input to the builder.

By default nodes provisioned by Server will use socket addresses starting with 127.0.1.1:9042, 127.0.1.2:9042, and so on. To alter this behavior you may pass in a custom AddressResolver by using Server.Builder.withAddressResolver(AddressResolver).

This behavior may also be altered by using withMultipleNodesPerIp(true). This mimics the behavior introduced in CASSANDRA-7544 which allows multiple Cassandra instances to run on the same IP, but with a different port.

Server is an AutoCloseable resource, thus can be used in a try-with-resources block or may be closed explicitly using close().

Provisioning Clusters and Nodes

With a Server instance in hand, you can provision Cluster and Nodes using Server.register.

register returns BoundCluster and BoundNode instances which each implement java.io.AutoCloseable and thus can be used in a try-with-resources block such that when the try block exits, the BoundCluster or BoundNode is automatically unregistered with the Server.

To register a single BoundNode:

import com.datastax.oss.simulacron.common.cluster.NodeSpec;
import com.datastax.oss.simulacron.server.BoundNode;

try (BoundNode node = server.register(NodeSpec.builder())) {
    // interact with Node
}

To register a Cluster:

import com.datastax.oss.simulacron.common.cluster.ClusterSpec;
import com.datastax.oss.simulacron.server.BoundCluster;

try (BoundCluster cluster = server.register(ClusterSpec.builder().withNodes(10,10))) {
    // interact with Cluster
}

Note that the input ClusterSpec and NodeSpec instances provided to register simply describe what simulacron should provision. The returned objects are 'bound' to the Server. Both Bound and Spec types share a base type, AbstractNode.

Cluster, DataCenter and Node configurability

ClusterSpec, DataCenterSpec and NodeSpecs can be configured in the following ways during construction with their builders:

All Subjects:

  • withCassandraVersion(String) - Configures the version of C* to be configured. Defaults to 3.0.12.
  • withDSEVersion(String) - Configures the version of DSE to be configured. If not present assumes not a DSE cluster. If present adds some extra columns to peers table to mimic a DSE node.
  • withName(String) - Gives a name to the subject. Is used as Cluster and DC name in metadata respectively.
  • withPeerInfo(String key, Object value) - Configures what peer columns should return. Value types must match what is expected or else defaults are used.
  • withPeerInfo(Map<String, Object> peerInfo) - Full version of peer info.

ClusterSpec:

  • withNodes(int nodeCount ...) - Convenience for specifying DataCenter layout of Cluster with each index of input representing a DC with number of nodes in that DC.

NodeSpec:

  • withAddress(SocketAddress address) - Configure the listening address for the to be constructed Node. If not provided, the configured AddressResolver will assign an address automatically.

In the general case, configuring everything at the ClusterSpec level is adequate, but in some cases you may want configure DataCenterSpec and NodeSpecs individually. DataCenterSpecs can be added to constructed ClusterSpec instances using addDataCenter(). NodeSpecs can be added to constructed DataCenterSpec instances using addNode(). For example:

import com.datastax.oss.simulacron.common.cluster.*;
import com.datastax.oss.simulacron.server.BoundCluster;
import com.datastax.oss.simulacron.server.Server;
import java.util.UUID;

Server server = Server.builder().build();

ClusterSpec cluster = ClusterSpec.builder().build();

// Add DC whose nodes are at C* 3.8.
DataCenterSpec dc = cluster.addDataCenter().withCassandraVersion("3.8").build();

// Add nodes to dc with their own configuration.
NodeSpec node0 = dc.addNode().withPeerInfo("rack", "rack2").build();
NodeSpec node1 = dc.addNode().withPeerInfo("rack", "rack1").withPeerInfo("host_id", UUID.randomUUID()).build();

// Nodes and DataCenters cannot be added after the Cluster is registered to the Server.
BoundCluster bCluster = server.register(cluster);

This API is admittedly non-ideal, it will be improved in the future.

Shortcuts for accessing DataCenters and Nodes

Both BoundCluster and BoundDataCenter provide convenience methods to quickly accessing their children objects respectively, i.e.:

try (BoundCluster cluster = server.register(ClusterSpec.builder().withNodes(5, 5, 5))) {
    // short cut to get a node in dc 0.
    BoundNode dc0node1 = cluster.node(1);
    // short cut to get dc 1.
    BoundNode dc1 = cluster.dc(1);
    // access dc 1, and then node 2 in that dc.
    BoundNode dc1node2 = dc1.node(2);
    // access node 3 in dc 2.
    BoundNode dc2node3 = cluster.dc(2, 3);
}

Driver Integration Support

Simulacron includes driver-xx (i.e. driver-3x) compatibility modules for various versions of the java driver. These modules are optional and merely provide convenience functionality.

SimulacronDriverSupport also provides defaultBuilder() methods for creating com.datastax.driver.core.Cluster.Builder instances that are preconfigured to communicate with simulacron BoundClusters.

import static com.datastax.oss.simulacron.driver.SimulacronDriverSupport.*;
import static com.datastax.oss.simulacron.common.cluster.*;

// Creates a builder that simply configures netty options such that it closes quickly.  This is useful
// for testing but has no intrinsic ties to simulacron functionality.
Cluster.Builder builder = defaultBuilder();

// Creates a builder that has its contact point pointing to the first node in the input cluster.
BoundCluster cluster = server.register(ClusterSpec.builder().withNodes(3));
Cluster.Builder builder2 = defaultBuilder(cluster);

// Creates a builder that has its contact point pointing to the the input node.
Cluster.Builder builder3 = defaultBuilder(cluster.node(1));

DriverTypeAdapters

DriverTypeAdapters offers methods for converting between types that share the same name and purpose. Both the driver and simulacron provide Consistency and WriteType implementations.

adapt() is meant to convert a driver type to a simulacron type. extract() is meant to convert a simulacron type to a driver type. For example:

import com.datastax.oss.driver.core.ConsistencyLevel;
import com.datastax.oss.driver.core.WriteType;
import static com.datastax.oss.simulacron.driver.DriverTypeAdapters.*;

// Convert from driver CL to simulacron CL
com.datastax.oss.simulacron.common.codec.ConsistencyLevel cl = adapt(ConsistencyLevel.ONE);

// Convert from simulacron CL to driver CL
ConsistencyLevel driverCl = extract(cl);

// Convert from driver WriteType to simulacron WriteType
com.datastax.oss.simulacron.common.codec.WriteType writeType = adapt(WriteType.SIMPLE);

// Convert from simulacron WriteType to simulacron WriteType
WriteType driverWriteType = extract(writeType);

Priming Responses

The priming API provides a mechanism for defining how simulacron-simulated nodes should respond to requests. Simulacron provides an easy to use fluent API, PrimeDsl, for priming queries.

A prime is broken up into two sections:

  • when: Defines the matching criteria, i.e.: When this query is made..
  • then: Defines what response to send, i.e.: Then send this response...

To begin a prime, simply use PrimeDsl.when to construct the when criteria, and then chain a then call with the desired result.

For example, the following primes the query select bar from foo to return a 'read timeout' error that specifies that 0 out of 1 responses were received for a query requiring ONE consistency level:

import com.datastax.oss.simulacron.common.stubbing.PrimeDsl.PrimeBuilder;
import static com.datastax.oss.simulacron.common.stubbing.PrimeDsl.*;

PrimeBuilder builder = when("select bar from foo").then(readTimeout(ConsistencyLevel.ONE, 0, 1, false));
Prime prime = builder.build();

To register the prime with a BoundCluster, BoundDataCenter or BoundNode simply call simply call prime.

node.prime(prime);

// as a short cut, you could simply pass in the builder, i.e.:
cluster.prime(
        when("select bar from foo")
        .then(readTimeout(ConsistencyLevel.ONE, 0, 1, false)));

Priming When (Requests)

PrimeDsl offers a few ways to prime when criteria:

  • when(String query): Creates a prime matching query string.
  • when(Request request): Creates a prime matching the input Request.

For more specific query criteria, the following can query methods can be used in conjunction with when:

  • query(String query, ConsistencyLevel consistency): Creates query criteria matching given query and consistency level.
  • query(String query, List<ConsistencyLevel> consistencies): Creates query criteria matching given query with any of the following consistencies.
  • query(String query, List<ConsistencyLevel> consistencies, Map<String, Object> params, Map<String, String> paramTypes): Creates query criteria matching the given query with any of the following consistencies, having the following parameter name, value mappings with the given name, value types.

In general, one will prime only queries. However one can go as far as to configure responses for all kinds of requests.

For example, it may be desired to prime a node such that it does not respond to Options messages. These are the messages that a driver may send as a means of doing heartbeats.

To prime this scenario:

import com.datastax.oss.simulacron.common.request.Options;
import static com.datastax.oss.simulacron.common.stubbing.PrimeDsl.*;

cluster.node(3).prime(when(Options.INSTANCE).then(noResult()));

Priming Then (Responses)

PrimeDsl offers a variety of response types that can be used as an input to then:

  • noRows(): A successful response with 0 rows and no metadata.
  • rows(): A successful response with the defined rows. See Priming Row Responses for more details.
  • alreadyExists(String keyspace, String table): An already exists exception for the given keyspace and table.
  • authenticationError(String message): An authentication error with the given message.
  • configurationError(String message): A configuration error with the given message.
  • closeConnection(DisconnectAction.Scope scope, CloseType closeType): Closes connections at the given scope with the given close type.
  • functionFailure(String keyspace, String function, List<String> argTypes, String detail): A function failure for the given function.
  • invalid(String message): An invalid error with the given message.
  • isBootstrapping(): An error that indicates that the server is still bootstrapping.
  • overloaded(String message): An overloaded error with the given message.
  • readFailure(ConsistencyLevel cl, int received, int blockFor, Map failureReasonByEndpoint, boolean dataPresent): A read failure error.
  • readTimeout(ConsistencyLevel cl, int received, int blockFor, boolean dataPresent): A read timeout error.
  • serverError(String message): A server error with the given message.
  • syntaxError(String message): A syntax error with the given message.
  • truncateError(String message): A truncate error with the given message.
  • unauthorized(String message): An unauthorized error with the given message.
  • unavailable(ConsistencyLevel cl, int required, int alive): An unavailable error.
  • unprepared(String message): An error indicating that the query was unprepared with the given message.
  • writeFailure(Consistencylevel cl, int required, int blockFor, Map failureReasonByEndpoint, WriteType writeType): A write failure error.
  • writeTimeout(ConsistencyLevel cl, int received, int blockFor, WriteType writeType): A write timeout error.
  • void\_(): A 'void' result, which is the same result as having no prime, but is useful if you want to configure delays.
  • noResult(): Indicates that queries matching this prime should return any response.

In addition, you may simply not provide a then. This indicates to simulacron to not respond to the given request.

Priming Row Responses

A specialized builder is available for priming row responses as this would be arduous otherwise. This is made available via PrimeDsl.rows.

The following primes a response to 'select bar,baz from foo' to return 2 rows.

It is expected that the type names in columnTypes map to real cql types. Every cql type (included nested collections and tuples) is supported with exception to UDTs.

import static com.datastax.oss.simulacron.common.stubbing.PrimeDsl.*;

cluster.node(1,2).prime(
        when("select bar,baz from foo")
        .then(rows()
            .row("bar", "hello", "baz", 72L)
            .row("bar", "world", "baz", 34L)
            .columnTypes("bar", "varchar", "baz", "bigint")
        ));

Priming Delay

It may be desired to delay a response being sent by the server, to do this, use PrimeBuilder.delay(long delay, TimeUnit delayUnit):

import static com.datastax.oss.simulacron.common.stubbing.PrimeDsl.*;

// delay response to query for 5 seconds.
cluster.prime(
        when("select bar,baz from foo")
        .then(noRows())
        .delay(5, TimeUnit.SECONDS)
);

Clearing Primes

Primes may be cleared by calling clearPrimes(boolean nested), i.e.:

// Clear primes for a node.
cluster.node(1).clearPrimes();

// Clear primes for a DC and all of its underlying nodes.
cluster.dc(1).clearPrimes(true);

// Clear primes, but only at the DC level.
cluster.dc(1).clearPrimes(false);

Using the Activity Log

By default simulacron will log all requests nodes received to an activity log for that node. To disable activity logging at the server level, one can simply do the following:

Server server = Server.builder()
  .withActivityLoggingEnabled(false)
  .build();

One may also disable activity logging at the cluster level when registering it:

BoundCluster cluster = server.register(ClusterSpec.builder().withNodes(5).build(),
  ServerOptions.builder().withActivityLoggingEnabled(false));

Accessing Activity Logs

To access logs for a BoundCluster, BoundDataCenter, or BoundServer simply call getLogs(), i.e.:

// get all query logs
List<QueryLog> logs = node.getLogs().getLogs();

// get only query logs for queries received that had a matching prime.
List<QueryLog> logs = node.getLogs(true).getLogs();

QueryLog provides a variety of metadata about each query made on that node.

Clearing Activity Logs

To clear activity logs, simply call clearLogs() on the target object you want to clear logs for, i.e.:

dc.clearLogs();

Accessing and Closing Connections

Simulacron offers away to retrieve the established socket connections by node and further a way to close those connections.

To access connections at a cluster, data center, or node level, call getConnections(), i.e.:

// cluster level
ClusterConnectionReport clusterReport = cluster.getConnections();
// dc level
DataCenterConnectionReport dcReport = cluster.dc(0).getConnections();
// node level
NodeConnectionReport nodeReport = cluster.node(0, 1).getConnections();

Depending on the subject, a ConnectionReport will be returned. These objects are hierarchical in a similar structure to cluster, data center, and node. NodeConnectionReport offers a getConnections() method which returns all established sockets to that node.

List<SocketAddress> connections = nodeReport.getConnections();

To simply retrieve the number of active connections, one could use getActiveConnections().

Closing Connections

closeConnections(CloseType closeType) offers a mechanism to close all connections for a given subject, i.e.:

import com.datastax.oss.simulacron.common.stubbing.CloseType;

// Request to close connections.
// The connections that were closed are returned.
ClusterConnectionReport report = cluster.closeConnections(CloseType.DISCONNECT);

CloseType offers a number of ways by which to close a connection:

  • DISCONNECT - Simply disconnects the connection.
  • SHUTDOWN_READ - Prevents the socket from being able to read any inbound data. Good for testing things like heartbeat failures.
  • SHUTDOWN_WRITE - Prevents the socket from being able to write any outbound data. This will typically close the connection so does not have much use over disconnect.

You can also close individual connections, i.e.:

// Close the earliest established connection to the node.
NodeConnectionReport nodeReport = cluster.node(0, 1).getConnections();
cluster.closeConnection(nodeReport.get(0), CloseType.SHUTDOWN_READ);

Disabling and Enabling Acceptance of Connections

Simulacron provides the capability to configure nodes to stop listening for new connections in addition to resuming acceptance of connections. This is useful for simulating node outages, or nodes being unresponsive for various reasons.

Disabling Connection Listener

Cluster, DataCenter and Node each offer a rejectConnections(int after, RejectScope rejectScope) method for telling nodes to disable listening for new connections, i.e.:

import com.datastax.oss.simulacron.server.RejectScope;

// simulate DC outage
cluster.dc(1).rejectConnections(0, RejectScope.STOP);

// alias for stopping immediately as done in previous instance
cluster.dc(1).stop();

// tell node to unbind listening for new connections after 1 successful connection while keeping existing connections open.
cluster.node(2).rejectConnections(1, RejectScope.UNBIND);

// tell cluster to continue accepting connections, but to not respond to 'STARTUP' requests.
// simulates scenario where network is responsive, but cassandra process is not.
cluster.rejectConnections(1, RejectScope.REJECT_STARTUP);

The after argument offers a way to tell simulacron to continue accepting connections until the provided number of connections have been established. This is useful for simulating partially initialized connection pools.

RejectScope offers a number of ways by which to reject connections:

  • UNBIND - Stops listening for new connections. Existing connections stay connected.
  • STOP - Meant to simulate the behavior of stopping a node. Closes existing connections and then unbinds to stop listening for new ones.
  • REJECT_STARTUP - Accepts new connections, but does not respond to 'startup' response. Existing connections stay connected. Meant to simulate behavior of when a cassandra process becomes unresponsive, but the host OS is still able to read off the socket.

Re-enabling Connection Listener

If node(s) previously had their connection listener disabled, they may be re-enabled by using acceptConnections(), i.e.:

// simulate DC outage
cluster.dc(1).stop();

// bring node 1 back up in dc 1.
cluster.node(1, 1).acceptConnections();

// start() is also provided as an alias to acceptConnections()
cluster.node(1, 1).start();

Note that if acceptConnections() is used on a subject that is already listening, there is no effect and the future completes immediately.

Async API

All methods that require network interactivity in Server, BoundCluster, BoundDataCenter and BoundNode have asynchronous counterpart methods that are used under the covers and are also exposed in the API. These methods have the same name as their synchronous version, but ending with Async. These methods return CompletionStage. For example:

// sync version
BoundCluster boundCluster = server.register(ClusterSpec.builder().withNodes(5));

// async version
CompletionStage<BoundCluster> future = server.registerAsync(ClusterSpec.builder().withNodes(5));