Multithreading and scaling

• 1: Multithreading concepts
• 2: Configuring multithreading
• 3: Optimizing multithreaded performance

1 - Multithreading concepts

This section is a brief overview on how AxoSyslog works in multithreaded mode. It is mainly for illustration purposes: the concept has been somewhat simplified and may not completely match reality.

AxoSyslog always uses multiple threads:

• A main thread that is always running

• A number of worker threads that process the messages. You can influence the behavior of worker threads using the threaded() option and the --worker-threads command-line option.

• Some other, special threads for internal functionalities. For example, certain destinations run in a separate thread, independently of the multithreading (threaded()) and --worker-threads settings of AxoSyslog.

The maximum number of worker threads AxoSyslog uses is the number of CPUs or cores in the host running AxoSyslog (up to 64). You can limit this value using the --worker-threads command-line option that sets the maximum total number of threads AxoSyslog can use, including the main AxoSyslog thread. However, the --worker-threads option does not affect the supervisor of AxoSyslog. The supervisor is a separate process (see The syslog-ng manual page), but certain operating systems might display it as a thread. In addition, certain destinations always run in a separate thread, independently of the multithreading (threaded()) and --worker-threads settings of AxoSyslog.

When an event requiring a new thread occurs (for example, AxoSyslog receives new messages, or a destination becomes available), AxoSyslog tries to start a new thread. If there are no free threads, the task waits until a thread finishes its task and becomes available. There are two types of worker threads:

• Reader threads read messages from a source (as many as possible, but limited by the log-fetch-limit() and log-iw-size() options). The thread then processes these messages, that is, performs filtering, rewriting and other tasks as necessary, and puts the log message into the queue of the destination. If the destination does not have a queue (for example, usertty), the reader thread sends the message to the destination, without the interaction of a separate writer thread.

• Writer threads take the messages from the queue of the destination and send them to the destination, that is, write the messages into a file, or send them to the syslog server over the network. The writer thread starts to process messages from the queue only if the destination is writable, and there are enough messages in the queue, as set in the flush-lines() option. Writer threads stop processing messages when the destination becomes unavailable, or there are no more messages in the queue.

Sources and destinations affected by multithreading

The following list describes which sources and destinations can use multiple threads. Changing the --worker-threads command-line option changes the number of threads available to these sources and destinations.

• The tcp and syslog(tcp) sources can process independent connections in separate threads. The number of independent connections is limited by the max-connections() option of the source. Separate sources are processed by separate thread, for example, if you have two separate tcp sources defined that receive messages on different IP addresses or port, AxoSyslog will use separate threads for these sources even if they both have only a single active connection.

• The udp, file, and pipe sources use a single thread for every source statement.

• The tcp, syslog, and pipe destinations use a single thread for every destination.

• The file destination uses a single thread for writing the destination file, but may use a separate thread for each destination file if the filename includes macros.

Sources and destinations not affected by multithreading

The following list describes sources and destinations that use a separate thread even if you disable multithreading in AxoSyslog, in addition to the limit set in the --worker-threads command-line option.

• The logstore destination uses separate threads for writing the messages from the journal to the logstore files, and also for timestamping. These threads are independent from the setting of the --worker-threads command-line option.

• Every sql destination uses its own thread. These threads are independent from the setting of the --worker-threads command-line option.

• The java destinations use one thread, even if there are multiple Java-based destinations configured. This thread is independent from the setting of the --worker-threads command-line option.

2 - Configuring multithreading

Starting with version 3.6, AxoSyslog runs in multithreaded mode by default. You can enable multithreading in AxoSyslog using the following methods:

• Globally using the threaded(yes) option.

• Separately for selected sources or destinations using the flags("threaded") option.

Example: Enabling multithreading

To enable multithreading globally, use the threaded option:

   options {
        threaded(yes) ;
    };

To enable multithreading only for a selected source or destination, use the flags("threaded") option:

   source s_tcp_syslog {
        network(
            ip(127.0.0.1)
            port(1999)
            flags("syslog-protocol", "threaded")
        );
    };

3 - Optimizing multithreaded performance

Destinations that have a queue process that queue in a single thread. Multiple sources can send messages to the same queue, so the queue can scale to multiple CPUs. However, when the writer thread writes the queue contents to the destination, it will be single-threaded.

Message parsing, rewrite rules, filters, and other types of message processing is performed by the reader thread in a sequential manner. This means that such operations can scale only if reading messages from the source can be multithreaded. For example, if a tcp source can process messages from different connections (clients) in separate threads. If the source cannot use multiple threads to process the messages, the operations will not scale.

To improve the processing power of AxoSyslog and scale to more processors, use the following methods:

• To improve scaling on the source side, use more sources, for example, more source files, or receive the messages from more parallel connections. For network sources, you can also configure a part of your clients to send the messages to a different port of your AxoSyslog server, and use separate source definitions for each port.

• On the destination side, when writing the log messages to files, use macros in the filename to split the messages to separate files (for example, using the ${HOST} macro). Files with macros in their filenames are processed in separate writer threads.

• On the destination side, when sending messages to an AxoSyslog server, you can use multiple connections to the server if you configure the AxoSyslog server to receive messages on multiple ports, and configure separate destinations on the clients to use both ports.