=head1 NAME
collectd-java - Documentation of collectd's "java plugin"
=head1 SYNOPSIS
LoadPlugin "java"
JVMArg "-verbose:jni"
JVMArg "-Djava.class.path=/opt/collectd/lib/collectd/bindings/java"
LoadPlugin "org.collectd.java.Foobar"
# To be parsed by the plugin
=head1 DESCRIPTION
The I plugin embeds a I (JVM) into I and
provides a Java interface to part of collectd's API. This makes it possible to
write additions to the daemon in Java.
This plugin is similar in nature to, but shares no code with, the I
plugin by Sebastian Harl, see L for details.
=head1 CONFIGURATION
A short outline of this plugin's configuration can be seen in L<"SYNOPSIS">
above. For a complete list of all configuration options and their semantics
please read L>.
=head1 OVERVIEW
When writing additions for collectd in Java, the underlying C base is mostly
hidden from you. All complex data types are converted to their Java counterparts
before they're passed to your functions. These Java classes reside in the
I namespace.
The I plugin will create one object of each class configured with the
B option. The constructor of this class can then register "callback
methods", i.Ee. methods that will be called by the daemon when
appropriate.
The available classes are:
=over 4
=item B
All API functions exported to Java are implemented as static functions of this
class. See L<"EXPORTED API FUNCTIONS"> below.
=item B
Corresponds to C, defined in F.
=item B
Corresponds to C, defined in F.
=item B
Corresponds to C, defined in F.
=item B
Corresponds to C, defined in F.
=item B
Corresponds to C, defined in F.
=item B
Corresponds to C, defined in F.
=back
In the remainder of this document, we'll use the short form of these names, for
example B. In order to be able to use these abbreviated names, you
need to B the classes.
=head1 EXPORTED API FUNCTIONS
All collectd API functions that are available to Java plugins are implemented
as Istatic> functions of the B class. This makes
calling these functions pretty straight forward. For example, to send an error
message to the daemon, you'd do something like this:
Collectd.logError ("That wasn't chicken!");
The following are the currently exported functions.
=head2 registerConfig
Signature: I B (I name,
I object);
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"config callback"> below.
=head2 registerInit
Signature: I B (I name,
I object);
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"init callback"> below.
=head2 registerRead
Signature: I B (I name,
I object)
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"read callback"> below.
=head2 registerWrite
Signature: I B (I name,
I object)
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"write callback"> below.
=head2 registerFlush
Signature: I B (I name,
I object)
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"flush callback"> below.
=head2 registerShutdown
Signature: I B (I name,
I object);
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"shutdown callback"> below.
=head2 registerLog
Signature: I B (I name,
I object);
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"log callback"> below.
=head2 registerNotification
Signature: I B (I name,
I object);
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"notification callback"> below.
=head2 registerMatch
Signature: I B (I name,
I object);
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"match callback"> below.
=head2 registerTarget
Signature: I B (I name,
I object);
Registers the B function of I with the daemon.
Returns zero upon success and non-zero when an error occurred.
See L<"target callback"> below.
=head2 dispatchValues
Signature: I B (I)
Passes the values represented by the B object to the
C function of the daemon. The "data set" (or list of
"data sources") associated with the object are ignored, because
C will automatically lookup the required data set. It
is therefore absolutely okay to leave this blank.
Returns zero upon success or non-zero upon failure.
=head2 getDS
Signature: I B (I)
Returns the appropriate I or B if the type is not defined.
=head2 logError
Signature: I B (I)
Sends a log message with severity B to the daemon.
=head2 logWarning
Signature: I B (I)
Sends a log message with severity B to the daemon.
=head2 logNotice
Signature: I B (I)
Sends a log message with severity B to the daemon.
=head2 logInfo
Signature: I B (I)
Sends a log message with severity B to the daemon.
=head2 logDebug
Signature: I B (I)
Sends a log message with severity B to the daemon.
=head1 REGISTERING CALLBACKS
When starting up, collectd creates an object of each configured class. The
constructor of this class should then register "callbacks" with the daemon,
using the appropriate static functions in B,
see L<"EXPORTED API FUNCTIONS"> above. To register a callback, the object being
passed to one of the register functions must implement an appropriate
interface, which are all in the B namespace.
A constructor may register any number of these callbacks, even none. An object
without callback methods is never actively called by collectd, but may still
call the exported API functions. One could, for example, start a new thread in
the constructor and dispatch (submit to the daemon) values asynchronously,
whenever one is available.
Each callback method is now explained in more detail:
=head2 config callback
Interface: B
Signature: I B (I ci)
This method is passed a B object, if both, method and
configuration, are available. B is the root of a tree representing
the configuration for this plugin. The root itself is the representation of the
BPluginE/E> block, so in next to all cases the children of the
root are the first interesting objects.
To signal success, this method has to return zero. Anything else will be
considered an error condition and the plugin will be disabled entirely.
See L<"registerConfig"> above.
=head2 init callback
Interface: B
Signature: I B ()
This method is called after the configuration has been handled. It is
supposed to set up the plugin. e.Eg. start threads, open connections, or
check if can do anything useful at all.
To signal success, this method has to return zero. Anything else will be
considered an error condition and the plugin will be disabled entirely.
See L<"registerInit"> above.
=head2 read callback
Interface: B
Signature: I B ()
This method is called periodically and is supposed to gather statistics in
whatever fashion. These statistics are represented as a B object and
sent to the daemon using L.
To signal success, this method has to return zero. Anything else will be
considered an error condition and cause an appropriate message to be logged.
Currently, returning non-zero does not have any other effects. In particular,
Java "read"-methods are not suspended for increasing intervals like C
"read"-functions.
See L<"registerRead"> above.
=head2 write callback
Interface: B
Signature: I B (I vl)
This method is called whenever a value is dispatched to the daemon. The
corresponding C "write"-functions are passed a C, so they can
decide which values are absolute values (gauge) and which are counter values.
To get the corresponding CDataSourceE>, call the B
method of the B object.
To signal success, this method has to return zero. Anything else will be
considered an error condition and cause an appropriate message to be logged.
See L<"registerWrite"> above.
=head2 flush callback
Interface: B
Signature: I B (I timeout, I identifier)
This method is called when the daemon received a flush command. This can either
be done using the C signal (see L) or using the I
plugin (see L).
If I is greater than zero, only values older than this number of
seconds should be flushed. To signal that all values should be flushed
regardless of age, this argument is set to a negative number.
The I specifies which value should be flushed. If it is not
possible to flush one specific value, flush all values. To signal that all
values should be flushed, this argument is set to I.
To signal success, this method has to return zero. Anything else will be
considered an error condition and cause an appropriate message to be logged.
See L<"registerFlush"> above.
=head2 shutdown callback
Interface: B
Signature: I B ()
This method is called when the daemon is shutting down. You should not rely on
the destructor to clean up behind the object but use this function instead.
To signal success, this method has to return zero. Anything else will be
considered an error condition and cause an appropriate message to be logged.
See L<"registerShutdown"> above.
=head2 log callback
Interface: B
Signature: I B (I severity, I message)
This callback can be used to receive log messages from the daemon.
The argument I is one of:
=over 4
=item *
org.collectd.api.Collectd.LOG_ERR
=item *
org.collectd.api.Collectd.LOG_WARNING
=item *
org.collectd.api.Collectd.LOG_NOTICE
=item *
org.collectd.api.Collectd.LOG_INFO
=item *
org.collectd.api.Collectd.LOG_DEBUG
=back
The function does not return any value.
See L<"registerLog"> above.
=head2 notification callback
Interface: B
Signature: I B (I n)
This callback can be used to receive notifications from the daemon.
To signal success, this method has to return zero. Anything else will be
considered an error condition and cause an appropriate message to be logged.
See L<"registerNotification"> above.
=head2 match callback
The match (and target, see L<"target callback"> below) callbacks work a bit
different from the other callbacks above: You don't register a match callback
with the daemon directly, but you register a function which, when called,
creates an appropriate object. The object creating the "match" objects is
called "match factory".
See L<"registerMatch"> above.
=head3 Factory object
Interface: B
Signature: I B
(I ci);
Called by the daemon to create "match" objects.
Returns: A new object which implements the B interface.
=head3 Match object
Interface: B
Signature: I B (I ds, I vl);
Called when processing a chain to determine whether or not a I
matches. How values are matches is up to the implementing class.
Has to return one of:
=over 4
=item *
B
=item *
B
=back
=head2 target callback
The target (and match, see L<"match callback"> above) callbacks work a bit
different from the other callbacks above: You don't register a target callback
with the daemon directly, but you register a function which, when called,
creates an appropriate object. The object creating the "target" objects is
called "target factory".
See L<"registerTarget"> above.
=head3 Factory object
Interface: B
Signature: I B
(I ci);
Called by the daemon to create "target" objects.
Returns: A new object which implements the B
interface.
=head3 Target object
Interface: B
Signature: I B (I ds, I vl);
Called when processing a chain to perform some action. The action performed is
up to the implementing class.
Has to return one of:
=over 4
=item *
B
=item *
B
=item *
B
=back
=head1 EXAMPLE
This short example demonstrates how to register a read callback with the
daemon:
import org.collectd.api.Collectd;
import org.collectd.api.ValueList;
import org.collectd.api.CollectdReadInterface;
public class Foobar implements CollectdReadInterface
{
public Foobar ()
{
Collectd.registerRead ("Foobar", this);
}
public int read ()
{
ValueList vl;
/* Do something... */
Collectd.dispatchValues (vl);
}
}
=head1 PLUGINS
The following plugins are implemented in I. Both, the B
option and the B block must be inside the
BPluginEjavaE> block (see above).
=head2 GenericJMX plugin
The GenericJMX plugin reads I (MBeans) from an I
using JMX. JMX is a generic framework to provide and query various management
information. The interface is used by Java processes to provide internal
statistics as well as by the I (JVM) to provide
information about the memory used, threads and so on.
The configuration of the I consists of two blocks: I
blocks that define a mapping of MBean attributes to the “types” used by
I, and I blocks which define the parameters needed to
connect to an I and what data to collect. The configuration of the
I is similar in nature, in case you know it.
=head3 MBean blocks
I blocks specify what data is retrieved from I and how that data
is mapped on the I data types. The block requires one string
argument, a name. This name is used in the I blocks (see below) to
refer to a specific I block. Therefore, the names must be unique.
The following options are recognized within I blocks:
=over 4
=item B I
Sets the pattern which is used to retrieve I from the I.
If more than one MBean is returned you should use the B option
(see below) to make the identifiers unique.
See also:
L
=item B I
Prefixes the generated I with I. I<(optional)>
=item B I
The I used by JMX to identify I include so called
I<“properties”> which are basically key-value-pairs. If the given object name
is not unique and multiple MBeans are returned, the values of those properties
usually differ. You can use this option to build the I from
the appropriate property values. This option is optional and may be repeated to
generate the I from multiple property values.
=item Bvalue /E> blocks
The I blocks map one or more attributes of an I to a value list
in I. There must be at least one Value block within each I
block.
=over 4
=item B type
Sets the data set used within I to handle the values of the I
attribute.
=item B I
Works like the option of the same name directly beneath the I block, but
sets the type instance instead. I<(optional)>
=item B I