Detecting Dead Connections
In this section we will discuss connection time-to-live (TTL) and explain how Apache ActiveMQ Artemis deals with crashed clients and clients which have exited without cleanly closing their resources.
Cleaning up Resources on the Server
Before an Apache ActiveMQ Artemis client application exits it is considered good
practice that it should close its resources in a controlled manner,
using a finally block.
Here's an example of a well behaved core client application closing its session and session factory in a finally block:
ServerLocator locator = null;
ClientSessionFactory sf = null;
ClientSession session = null;
try {
locator = ActiveMQClient.createServerLocatorWithoutHA(..);
sf = locator.createClientSessionFactory();;
session = sf.createSession(...);
... do some stuff with the session...
} finally {
if (session != null) {
session.close();
}
if (sf != null) {
sf.close();
}
if(locator != null) {
locator.close();
}
}
And here's an example of a well behaved JMS client application:
Connection jmsConnection = null;
try {
ConnectionFactory jmsConnectionFactory = new ActiveMQConnectionFactory("tcp://localhost:61616");
jmsConnection = jmsConnectionFactory.createConnection();
... do some stuff with the connection...
} finally {
if (connection != null) {
connection.close();
}
}
Or with using auto-closeable feature from Java, which can save a few lines of code:
try (
ActiveMQConnectionFactory jmsConnectionFactory = new ActiveMQConnectionFactory("tcp://localhost:61616");
Connection jmsConnection = jmsConnectionFactory.createConnection()) {
... do some stuff with the connection...
}
Unfortunately users don't always write well behaved applications, and sometimes clients just crash so they don't have a chance to clean up their resources!
If this occurs then it can leave server side resources, like sessions, hanging on the server. If these were not removed they would cause a resource leak on the server and over time this result in the server running out of memory or other resources.
We have to balance the requirement for cleaning up dead client resources with the fact that sometimes the network between the client and the server can fail and then come back, allowing the client to reconnect. Apache ActiveMQ Artemis supports client reconnection, so we don't want to clean up "dead" server side resources too soon or this will prevent any client from reconnecting, as it won't be able to find its old sessions on the server.
Apache ActiveMQ Artemis makes all of this configurable via a connection TTL. Basically, the TTL determines how long the server will keep a connection alive in the absence of any data arriving from the client. The client will automatically send "ping" packets periodically to prevent the server from closing it down. If the server doesn't receive any packets on a connection for the connection TTL time, then it will automatically close all the sessions on the server that relate to that connection.
The connection TTL is configured on the URI using the connectionTtl
parameter.
The default value for connection ttl on an "unreliable" connection (e.g.
a Netty connection using the tcp URL scheme) is 60000ms, i.e. 1 minute.
The default value for connection ttl on a "reliable" connection (e.g. an
in-vm connection using the vm URL scheme) is -1. A value of -1 for
connectionTTL means the server will never time out the connection on
the server side.
If you do not wish clients to be able to specify their own connection
TTL, you can override all values used by a global value set on the
server side. This can be done by specifying the
connection-ttl-override attribute in the server side configuration.
The default value for connection-ttl-override is -1 which means "do
not override" (i.e. let clients use their own values).
The logic to check connections for TTL violations runs periodically on
the broker. By default, the checks are done every 2,000 milliseconds.
However, this can be changed if necessary by using the
connection-ttl-check-interval attribute.
Closing Forgotten Resources
As previously discussed, it's important that all core client sessions
and JMS connections are always closed explicitly in a finally block
when you are finished using them.
If you fail to do so, Apache ActiveMQ Artemis will detect this at garbage collection time, and log a warning (If you are using JMS the warning will involve a JMS connection).
Apache ActiveMQ Artemis will then close the connection / client session for you.
Note that the log will also tell you the exact line of your user code where you created the JMS connection / client session that you later did not close. This will enable you to pinpoint the error in your code and correct it appropriately.
Detecting Failure from the Client
In the previous section we discussed how the client sends pings to the server and how "dead" connection resources are cleaned up by the server. There's also another reason for pinging, and that's for the client to be able to detect that the server or network has failed.
As long as the client is receiving data from the server it will consider the connection to be still alive.
If the client does not receive any packets for a configurable number
of milliseconds then it will consider the connection failed and will
either initiate failover, or call any FailureListener instances (or
ExceptionListener instances if you are using JMS) depending on how
it has been configured.
This is controlled by setting the clientFailureCheckPeriod parameter
on the URI your client is using to connect, e.g.
tcp://localhost:61616?clientFailureCheckPeriod=30000.
The default value for client failure check period on an "unreliable"
connection (e.g. a Netty connection) is 30000 ms, i.e. 30 seconds. The
default value for client failure check period on a "reliable" connection
(e.g. an in-vm connection) is -1. A value of -1 means the client
will never fail the connection on the client side if no data is received
from the server. Typically this is much lower than connection TTL to
allow clients to reconnect in case of transitory failure.
Configuring Asynchronous Connection Execution
Most packets received on the server side are executed on the remoting thread. These packets represent short-running operations and are always executed on the remoting thread for performance reasons.
However, by default some kinds of packets are executed using a thread from a thread pool so that the remoting thread is not tied up for too long. Please note that processing operations asynchronously on another thread adds a little more latency. These packets are:
org.apache.activemq.artemis.core.protocol.core.impl.wireformat.RollbackMessageorg.apache.activemq.artemis.core.protocol.core.impl.wireformat.SessionCloseMessageorg.apache.activemq.artemis.core.protocol.core.impl.wireformat.SessionCommitMessageorg.apache.activemq.artemis.core.protocol.core.impl.wireformat.SessionXACommitMessageorg.apache.activemq.artemis.core.protocol.core.impl.wireformat.SessionXAPrepareMessageorg.apache.activemq.artemis.core.protocol.core.impl.wireformat.SessionXARollbackMessage
To disable asynchronous connection execution, set the parameter
async-connection-execution-enabled in broker.xml to
false (default value is true).