[syslog-ng] syslog-ng-3.8.1 has been released

Laszlo Budai laszlo.budai at outlook.com
Fri Aug 19 13:47:54 CEST 2016

Library updates

  *   Kafka-client updated to version to
  *   Minimal required version of hiredis is set to 0.11.0 to avoid possible deadlocks
  *   Minimal version of libdbi is set to 0.9.0

New dependencies

  *   From now autoconf-archive package is a build-dependency.

Improvements and features

  *   Added the long-waited disk-buffer.
  *   date-parser ported from incubator to upstream
  *   New template functions: min, max, sum, average
  *   Added Apache-accesslog-parser
  *   Added loggly destination
  *   Added logmatic destination
  *   Added template function for supporting CEF.
  *   cURL-based HTTP destination driver added (implemented in C programming language)
  *   SELinux policy installer script now has support for Red Hat Enterprise Linux/CentOS/ Oracle Linux 5, 6 and 7.
  *   Implemented add-contextual-data: With add-context-data syslog-ng can use an external database file to append custom name-value pairs on incoming logs (to enrich messages). The 'database' is actually a file that containing <selector, name, value> records. Currently only CSVformat is supported.
It is like geoip parser where the selector is $HOST, but in this case, the user can define the selector, and also the database contents.


  *   Program destination/source drivers

     *   Added inherit-environment configuration option to program source and destination. When it is set to true then the process will inherit the entire environment of the parent process.
     *   Added keep-alive option to program destination (afprog). This option will control whether the destination program should be terminated at reload, or should be left running.
  *   Java drivers

     *   HTTP destination
        *   Added the ability to use templates in both url and message.
     *   ElasticSearch Destination driver :
        *   Support 2.2.x series of ElasticSearch (transport and node mode) .
        *   Support Shield plugin for both ElasticSearch 1.x and ElasticSearch 2.x .
        *   Implemented new mode (HTTP) that can work with ElasticSearch 1.x, ElasticSearch 2.x, and even with Elastic 5. HTTP mode is based on a Java HTTP Rest client (Jest :https://github.com/searchbox-io/Jest). Note: make install will copy Jest library to the syslog-ng install directory.
  *   MongoDB destination driver

     *   Replaced submodule limongo-client with mongo-c-driver.
     *   Additional support for previous syntax used by libmongo-client before we started using mongo-c-driver and its URI syntax exclusively. Note that these are plainly translated to a connection URI without much sanity checking or preserving their former semantic meaning. So various aspects of the MongoDB connection like health checks, retries, error reporting and synchronicity will still follow the slightly altered semantics of mongo-c-driver.
  *   Riemann destination driver

     *   Use cert-file() and key-file() options to match afsocket keywords as the same way as afsocket drivers use these options. The old one still work though.

Rewrite rules

  *   Introduced template options in rewrite rules.
  *   Added unset operation to make it possible to unset a specific name-value pair for a logmessage.


  *   kvformat: make it possible to specify name-value separator
  *   linux-audit-scanner: recognize a0-a9* as fields to be decoded Argument lists are encoded in a0, a1, ... fields that can potentially be hex-encoded.
  *   csv-parser has been refactored, extended with new dialect and prefix options. Dialect is to convey CSV format information, instead of using flags Prefix option gets prefixed to all column names, just like with other parsers.


  *   added groupingby() parser that can perform simple correlation on
log messages. In a way it is similar to the SQL GROUP BY operation, where
an aggregate of a set of input records can be calculated.
The major difference between SQL GROUP BY and groupingby() is that the
first always operates on a enumerable list of records, whereas
groupingby() works on a stream of data.
A few use-cases where this can be useful:

     *   Linux audit logs
     *   postfix logs
  *   added create-context action

Added a new possible action in the element, to create
a new correlation context out of the current message and its associated
context. This can be used to "split" a state.

  *   Added NLSTRING parser that captures a string until the
following end-of-line. It can be used in patterns as: @NLSTRING:value@
It doesn't expect any additional parameters. This makes it pretty easy to
parse multi-line Windows logs.

Miscellaneous features

  *   syslog-debun (debug bundle script for syslog-ng) has been improved


  *   geoip-parser: When default database if not specified, syslog-ng crashed.
  *   Added support for multiple drivers with the same name in syslog-ng config.
  *   Fixed aack counting logic for junctions that have branches that modify the LogMessage.
  *   Fixed a potential crash for code that uses log_msg_clear() in production (e.g. syslog-parser()).
  *   Fixed potential crash in reload logic
  *   system(): use string comparison instead of numeric in PID rewrite The meaning of the != operator has been fixed to refer to numeric comparison in @version<https://github.com/version>: 3.8, so make sure we are using string comparison.
  *   Support encoding on glib compiled with libiconv
  *   pdbtool: Fix the ordering of the debug-info list in PatternDB
  *   afprog: Don't kill our own process group
If, for some reason, the program source/destination failed to set up its own process group, we need to make sure we do not run killpg() on that process group, as it would kill ourselves.
  *   Handle option names with hyphen (-) characters in java scls
  *   dnscache performance improved Instead of getting rid off the per-thread DNSCache when a worker thread exits, store them in a linked list and acquire them as a new thread starts. The set of cached hostnames are valuable as worker threads come and go (they exit after 10seconds of inactivity), but without this reusing of cache instances, our DNS cache is filled again and again.
  *   Fixed IPv6 parser in patterndb.
  *   Fixed journald program name flapping
  *   Fixed create-dirs() inheritance in file destinations
  *   Fixed pass-unix-credentials() global inheritance in afunix The global pass-unix-credentialsoption was not inherited in afunix-source if the options{}; block was positioned lower in the configuration file than the given module declaration.
  *   Fixed create-dirs() global inheritance in afunix When the global create-dirs option was set toyes, the local one was ignored.
  *   Fixed byteorder handling on bigendian systems in netmask6 filter
  *   Fixed flow-control issue when overflow queue is full (suspending source by setting the window size to 0).
  *   Log HTTP response error codes in HTTPDestination (Java).
  *   Fixed potential leaks related $(sanitize) argument parsing in basicfuncs.
  *   Fixed a memory leak in python debugger
  *   Fixed a use-after-free bug in templates.
  *   Fixed a memory leak around reload in netmask6 filter.
  *   Fixed a memory leak in LogProtoBufferedServer in case the encoding() option is used.
  *   configure: don't override $enable_python while executing pkg-config
  *   Fixed BSD timestamp parsing in syslog-format.
  *   Fixed a SIGPIPE bug in program destination.
  *   Error handling has been improved in AMQP destination.
  *   value-pairs performance improvements, memleak fixes
  *   Various issues around UTF-8 support fixed.
  *   Fixed integer overflow in numerical operations template function
  *   Fixed an integer underflow in afsocket.
  *   Fixed numerical comperisons issues around filters. There's a problem in straight fixing this issue though: anyone who used the numeric operators erroneously will have their behaviour changed, therefore this patch also adds a configuration update warning in case someone is using the wrong syntax.
  *   Fixed kernel log message time drift on Linux.
  *   Take CRLF sequences equivalent to an LF in patterndb.
Windows logs contain embedded CRLFs which is difficult to match against from db-parser(), as we use a UNIX text file to store the patterns. Also, the fact that the input contains CRLF whereas our patterns only contain an LF makes it a very unintuitive non-match, which is difficult to debug.
  *   When syslog-ng failed to insert data into Redis, it has crashed.
  *   When device file is set as a file destination then syslog-ng will not try to change the permission of the device file.
  *   Various fixes around config file parsing:
     *   in some circumstances syslog-ng crashed when the config file contained non-readable characters
     *   fixed a memory leak
     *   fixed memory leak around backtick substitution

Notes to the Developers

  *   copyright cleanup in source tree

  *   install tools and scl under a syslog-ng specific subdirectory

These should never be installed in /usr/share directly, but rather under a
subdirectory and as described in

we should do that right within the source and not rely on packaging tools
to do it for us.This will trigger a required change in packaging scripts to
avoid changing the --datadir, as the default of
/usr/share should work out-of-the-box.

  *   Support for native-lanugage (compiled languages, like Rust) bindings.
These bindings just forward the calls to the native side.

This whole module compiles into a static library
(libsyslog-ng-native-connector.a) which is linked to all external native
modules. A native module defines the required functions
(like native_parser_proxy_new()) so those symbols will be resolved.

Some symbols have the visibility(hidden) attribute applied to them. Those
symbols are defined by the other half of the native bindings, we only need
their signature here. They are hidden because their definition is contained
in other source files but we would like to keep them as "library private".
If they are exported, the dynamic linker will resolve them when a module is
loaded, therefore every module would be mapped to the first loaded one.
It is best to hide everything in this native connector.

  *   Support system librabbitmq-c (AMQP destination)

Currently only the internal librabbitmq-c is supported,
if we want to use the preexisting library, the configuration
will fail.
This change is required if we want to get rid of the
internal libraries.

  *   Added check-valgrind target to Makefile

  *   Remove dependency on Gradle for building Java language bindings
(not modules, just the language binding)

  *   Experimental CMake support added

  *   Experimental OSX support added
  *   Improved Travis build matrix
  *   Added plugin skeleton creator.
  *   Debian packaging improved: Debian packaging from unofficial OBS repository has been ported to upstream. From now, someone could easily build debian packages from upstream source.


syslog-ng is developed as a community project, and as such it relies
on volunteers, to do the work necessarily to produce syslog-ng.

Reporting bugs, testing changes, writing code or simply providing
feedback are all important contributions, so please if you are a user
of syslog-ng, contribute.

We would like to thank the following people for their contribution:

Adam Istvan Mozes, Andras Mitzki, Arnaud Vamorec, Balazs Scheidler,
David Schweikert, Fabien Wernli, Flavio Medeiros, Hanno Böck,
Henrik Grindal Bakken, Gergo Nagy, Gyorgy Pasztor, Laszlo Budai,
Laszlo Varady, Marc Falzon, Noemi Vanyi, Peter Czanik, Robert Fekete,
Tamas Nagy,Tibor Benke, Viktor Juhasz, Viktor Tusa, Vincent Bernat,
Zdenek Styblik, Zoltan Fried, Zoltan Pallagi, Yilin Li




Laszlo Budai

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.balabit.hu/pipermail/syslog-ng/attachments/20160819/83820ff1/attachment-0001.htm 

More information about the syslog-ng mailing list