12.10. The Python Parser

The Python Log Parser (available in syslog-ng OSE version 3.10 and later) allows you to write your own parser in Python. Practically, that way you can process the log message (or parts of the log message) any way you need. For example, you can import external Python modules to process the messages, query databases to enrich the messages with additional data, and many other things.

  • Currently only Python 2.7 is supported.

  • The Python block must be a top-level block in the syslog-ng OSE configuration file. If you store the Python code in a separate Python file and only include it in the syslog-ng OSE configuration file, make sure that the PYTHON_PATH environment variable includes the path to the Python file, and export the PYTHON_PATH environment variable. For example: export PYTHONPATH=/opt/syslog-ng/etc

  • The Python object is initiated only once, when syslog-ng OSE is started or reloaded. That means it keeps the state of internal variables while syslog-ng OSE is running.

  • The Python block can contain multiple Python functions.

  • Using Python code in syslog-ng OSE can significantly decrease the performance of syslog-ng OSE, especially if the Python code is slow.

  • Validate and lint the Python code before using it. The syslog-ng OSE application does not do any of this.

Declaration: 

Python parsers consist of two parts. The first is a syslog-ng OSE parser object that you use in your syslog-ng OSE configuration, for example, in the log path. This parser references a Python class, which is the second part of the Python parsers. The Python class processes the log messages it receives, and can do virtually anything that you can code in Python.

parser <name_of_the_python_parser>{
  python(
    class("<name_of_the_python_class_executed_by_the_parser>")
  );
};

python {
import re
class MyParser(object):
    def init(self, options):
        '''Optional. This method is executed when syslog-ng is started or reloaded.'''
        return True
    def deinit(self):
        '''Optional. This method is executed when syslog-ng is stopped or reloaded.'''
        return True
    def parse(self, msg):
        '''Required. This method receives and processes the log message.'''
        return True
};

Methods of the python() parser

The init (self, options) method (optional). 

The syslog-ng OSE application initializes Python objects only when it is started or reloaded. That means it keeps the state of internal variables while syslog-ng OSE is running. The init method is executed as part of the initialization. You can perform any initialization steps that are necessary for your parser to work. For example, if you want to perform a lookup from a file or a database, you can open the file or connect to the database here, or you can initialize a counter that you will increase in the parse() method.

The return value of the init() method must be True. If it returns False, or raises an exception, syslog-ng OSE will not start.

options: This optional argument contains the contents of the options() parameter of the parser object as a Python dict.

parser my_python_parser{
  python(
    class("MyParser")
    options("regex", "seq: (?P<seq>\\d+), thread: (?P<thread>\\d+), runid: (?P<runid>\\d+), stamp: (?P<stamp>[^ ]+) (?P<padding>.*$)")
  );
};
class MyParser(object):
    def init(self, options):
        pattern = options["regex"]
        self.regex = re.compile(pattern)
        self.counter = 0
        return True

The parse(self, log_message) method. 

The parse() method processes the log messages it receives, and can do virtually anything that you can code in Python. This method is required, otherwise syslog-ng OSE will not start.

The return value of the parse() method must be True. If it returns False, or raises an exception, syslog-ng OSE will drop the message.

  • To reference a name-value pair or a macro in the Python code, use the following format. For example, if the first argument in the definition of the function is called log-message, the value of the HOST macro is log-message['HOST'], and so on. (The log-message contains the entire log message (not just the text body) in a structure similar to a Python dict, but it is actually an object.)

  • You can define new name-value pairs in the Python function. For example, if the first argument in the definition of the function is called log-message, you can create a new name-value pair like this: log_message["new-macro-name"]="value". This is useful when you parse a part of the message from Python, or lookup a value based on data extracted from the log message.

    Note that the names of the name-value pairs are case-sensitive. If you create a new name-value pair called new-macro-name in Python, and want to reference it in another part of the syslog-ng OSE configuration file (for example, in a template), use the ${new-macro-name} macro.

  • You cannot override hard macros (see Section 11.1.4, Hard vs. soft macros).

  • To list all available keys (names of name-value pairs), use the log_message.keys() function.

The deinit(self) method (optional). 

This method is executed when syslog-ng OSE is stopped or reloaded.

Warning

It is common practice for log rotate solutions to reload syslog-ng OSE (by sending a HUP signal or using the operating system's init subsystem) and for users to execute syslog-ng-ctl reload (to start a configuration file reload). Care should be taken in these cases, because the methods and attributes defined in a Python parser block definition lose their context and state during a syslog-ng OSE reload.

Example: Parse loggen logs

The following sample code parses the messages of the loggen tool (for details, see loggen(1)). The following is a sample loggen message:

<38>2017-04-05T12:16:46 localhost prg00000[1234]: seq: 0000000000, thread: 0000, runid: 1491387406, stamp: 2017-04-05T12:16:46 PADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADDPADD

The syslog-ng OSE parser object references the LoggenParser class and passes a set of regular expressions to parse the loggen messages. The init() method of the LoggenParser class compiles these expressions into a pattern. The parse method uses these patterns to extract the fields of the message into name-value pairs. The destination template of the syslog-ng OSE log statement uses the extracted fields to format the output message.

@version: 3.12
@include "scl.conf"
parser my_python_parser{
  python(
    class("LoggenParser")
    options("regex", "seq: (?P<seq>\\d+), thread: (?P<thread>\\d+), runid: (?P<runid>\\d+), stamp: (?P<stamp>[^ ]+) (?P<padding>.*$)")
  );
};
log {
  source { tcp(port(5555)); };
  parser(my_python_parser);
  destination {  file("/tmp/regexparser.log.txt" template("seq: $seq thread: $thread runid: $runid stamp: $stamp my_counter: $MY_COUNTER"));};
};
python {
import re
class LoggenParser(object):
    def init(self, options):
        pattern = options["regex"]
        self.regex = re.compile(pattern)
        self.counter = 0
        return True
    def deinit(self):
        return True
    def parse(self, log_message):
        match = self.regex.match(log_message['MESSAGE'])
        if match:
            for key, value in match.groupdict().items():
                log_message[key] = value
            log_message['MY_COUNTER'] = self.counter
            self.counter += 1
            return True
        return False
};

Example: Parse Windows eventlogs in Python - performance

The following example uses regular expressions to process Windows log messages received in XML format from the syslog-ng Agent for Windows application. The parser extracts different fields from messages received from the Security and the Application eventlog containers. Using the following configuration file, syslog-ng OSE could process about 25000 real-life Windows log messages per second.

@version: 3.12
options {
    keep_hostname(yes);
    keep_timestamp(no);
    stats_level(2);
    use_dns(no);
};
source s_network_aa5fdf25c39d4017a8e504cdb641b477 {
    network(flags(no-parse)
        ip(0.0.0.0)
        log_fetch_limit(1000)
        log_iw_size(100000)
        max_connections(100)
        port(514));
};
parser p_python_parser_79c31da44bb64de6b5de84be4ae15a15 {
    python(options("regex_for_security", ".* Security ID:  (?P<security_id>\\S+)   Account Name:  (?P<account_name>\\S+)   Account Domain:  (?P<account_domain>\\S+)   Logon ID:  (?P<logon_id>\\S+).*Process Name: (?P<process_name>\\S+).*EventID (?P<event_id>\\d+)", "regex_others", "(.*)EventID (?P<event_id>\\d+)")
class("EventlogParser"));
};
destination d_file_78363e1dd90c4ebcbb0ee1eff5a2e310 {
    file("/var/testdb_working_dir/fcd713a2-d48e-4025-9192-ec4a9852cafa.$HOST"
        flush_lines(1000)
        log_fifo_size(200000));
};
log {
    source(s_network_aa5fdf25c39d4017a8e504cdb641b477);
    parser(p_python_parser_79c31da44bb64de6b5de84be4ae15a15);
    destination(d_file_78363e1dd90c4ebcbb0ee1eff5a2e310);

    flags(flow-control);
};

python {
import re
class EventlogParser(object):
    def init(self, options):
        self.regex_security = re.compile(options["regex_for_security"])
        self.regex_others = re.compile(options["regex_others"])
        return True
    def deinit(self):
        return True
    def parse(self, log_message):
        security_match = self.regex_security.match(log_message['MESSAGE'])
        if security_match:
            for key, value in security_match.groupdict().items():
                log_message[key] = value
        else:
            others_match = self.regex_others.match(log_message['MESSAGE'])
            if others_match:
                for key, value in others_match.groupdict().items():
                    log_message[key] = value
        return True
};