syslog-ng 3.4.2 has been released
------------------------------------------------------------------------------ PACKAGE : syslog-ng VERSION : 3.4.2 SUMMARY : new stable release DATE : Jun 3, 2013 ------------------------------------------------------------------------------ DESCRIPTION: A new stable version of syslog-ng Open Source Edition (3.4.2) has been released. For latest fixes in the 3.4.x feature branch you are recommended to upgrade to this version. CHANGES: 3.4.2 Mon, 3 Jun 2013 16:00:00 +0100 This is the first bug-fix release for the 3.4.x series, collecting significant fixes over the 3.4.1 release. Upgrading to this release is highly recommended. Bugfixes ======== * loggen will bail out earily if using Unix sockets without a path specified. * Using the same filter for multiple destinations does not trigger a crash anymore. [#219] * An issue with log branch propagation has been fixed, restoring compatibility with the 3.3 branch (this mostly affected flags(final) use cases). [#220] * The systemd unit file now specified Restart=on-failure. [#222] * Support for empty blocks in the configuration were restored, syslog-ng does not crash anymore when encountering one. * An assertion was fixed in patterndb, that prevented one from overriding built-in values, such as HOST. [#225] * The persist state file (syslog-ng.persist) is now marked close-on-exec, so it does not leak through to forked subprocesses. * The system() source now treats /proc/kmsg as optional: if it is not readable to the user running syslog-ng, it will not be included. * Proper escape handling in backtick insertion has been implemented, so string arguments to block parameters are - again - treated correctly. [#235] * A rare race condition in the SQL, MongoDB, SMTP and AMQP destinations have been fixed. * A possible deadlock was fixed within the suppress() handling code. Credits ======= Attila M. Magyar <athos@balabit.hu> Balazs Scheidler <bazsi@balabit.hu> Evan Rempel <erempel@uvic.ca> Gergely Nagy <algernon@balabit.hu> Paul Dann <pdgiddie+balabit@gmail.com> Peter Czanik <czanik@balabit.hu> Tamas Pal <folti@balabit.hu> DOWNLOAD: You can download the source or binary packages from: http://www.balabit.com/network-security/syslog-ng/opensource-logging-system/... The documentation of the syslog-ng Open Source Edition is available in The syslog-ng Open Source Edition Administrator's Guide at http://www.balabit.com/support/documentation/
Deployed version 3.4.2 into our infrastructure and noticed the CPU profile change depicted in this image. The change from 3.4.1 to 3.4.2 occurred at 09:19 Looks like about 30% more user space CPU and 50% more system CPU.
On 06/03/2013 07:07 PM, Evan Rempel wrote:
Deployed version 3.4.2 into our infrastructure and noticed the CPU profile change depicted in this image.
The change from 3.4.1 to 3.4.2 occurred at 09:19
Looks like about 30% more user space CPU and 50% more system CPU.
Strange. While stress testing 3.4.2 before release I had quite the opposite experience. No nice graphs to include, I just watched the XFCE CPU graph applet, and found it used ~50% less CPU than with 3.4.1 and the same configuration (standard openSUSE syslog-ng config with network source and destination added). Bye, -- Peter Czanik (CzP) <czanik@balabit.hu> BalaBit IT Security / syslog-ng upstream http://czanik.blogs.balabit.com/
Try adding a program and/or a pipe destination. Another post to the list indicated that by removing the pipe destination, CPU returned to expected levels, and I use both program and pipe destinations. Evan. On 06/05/2013 02:16 AM, Peter Czanik wrote:
On 06/03/2013 07:07 PM, Evan Rempel wrote:
Deployed version 3.4.2 into our infrastructure and noticed the CPU profile change depicted in this image.
The change from 3.4.1 to 3.4.2 occurred at 09:19
Looks like about 30% more user space CPU and 50% more system CPU.
Strange. While stress testing 3.4.2 before release I had quite the opposite experience. No nice graphs to include, I just watched the XFCE CPU graph applet, and found it used ~50% less CPU than with 3.4.1 and the same configuration (standard openSUSE syslog-ng config with network source and destination added). Bye,
-- Evan Rempel erempel@uvic.ca Senior Systems Administrator 250.721.7691 Data Centre Services, University Systems, University of Victoria
Any chance of doing a perf record -g and perf report on.the production system? I can't remember anything particular. However Algernon should know better. He created this release. On Jun 5, 2013 11:16 AM, "Peter Czanik" <czanik@balabit.hu> wrote:
On 06/03/2013 07:07 PM, Evan Rempel wrote:
Deployed version 3.4.2 into our infrastructure and noticed the CPU profile change depicted in this image.
The change from 3.4.1 to 3.4.2 occurred at 09:19
Looks like about 30% more user space CPU and 50% more system CPU.
Strange. While stress testing 3.4.2 before release I had quite the opposite experience. No nice graphs to include, I just watched the XFCE CPU graph applet, and found it used ~50% less CPU than with 3.4.1 and the same configuration (standard openSUSE syslog-ng config with network source and destination added). Bye,
-- Peter Czanik (CzP) <czanik@balabit.hu> BalaBit IT Security / syslog-ng upstream http://czanik.blogs.balabit.com/
______________________________________________________________________________ Member info: https://lists.balabit.hu/mailman/listinfo/syslog-ng Documentation: http://www.balabit.com/support/documentation/?product=syslog-ng FAQ: http://www.balabit.com/wiki/syslog-ng-faq
After building/installing version 3.4.2 I am seeing many more segmentation faults. I used to get 1-2 per week, and after the update the errors are occurring more than 1 every 30 minutes. As a side note, I got an eventlog package of 0.2.13 from Balabit some time ago, and have never seen it available again. the current release of eventlog seems to be eventlog_0.2.12+20120504+1700.tar.gz but software with release versions that include dates alway make me nervous. Would the developers expect the eventlog versions to be so tightly coupled with the release version of syslog-ng that a mismatch would result in segmentation faults? Evan.
Evan Rempel <erempel@uvic.ca> writes:
After building/installing version 3.4.2 I am seeing many more segmentation faults. I used to get 1-2 per week, and after the update the errors are occurring more than 1 every 30 minutes.
Would it be possible to get a backtrace of those segfaults?
As a side note, I got an eventlog package of 0.2.13 from Balabit some time ago, and have never seen it available again. the current release of eventlog seems to be
eventlog_0.2.12+20120504+1700.tar.gz
but software with release versions that include dates alway make me nervous.
Unfortunately, my knowledge of our internal build & release system is sadly lacking, and can't really see how I could bundle the correct eventlog version with syslog-ng releases... But 0.2.12 and anything newer should work just fine.
Would the developers expect the eventlog versions to be so tightly coupled with the release version of syslog-ng that a mismatch would result in segmentation faults?
Nope, that is very unlikely. -- |8]
participants (5)
-
Balazs Scheidler
-
devel＠balabit.hu
-
Evan Rempel
-
Gergely Nagy
-
Peter Czanik