February 2022 - syslog-ng - lists.balabit.hu

UDP drops with syslog-ng 3.24.1-1
by brian hoffman 05 Oct '22

05 Oct '22

We've been centrally logging with syslog-ng for about 5 years now. Over that time, the number of sources has grown significantly, and at some point we crossed a line where drops were happening (a quick survey of 3 million syslog packets yielded 420 unique currently sending hosts). After much research and experimentation, we've been able to get to the point where throughout the day there are 0 drops for the most part. This was achieved by installing the latest syslog-ng (not the RedHat packaged one) and creating a source for each CPU. Occasionally, though, we still have periods of drops so I'm trying to eliminate these last few. Here are some relevant configuration items:2 RedHat 7.8 VMs (load balanced via an F5) with 16GB memory and 4 CPUs each running syslog-ng-3.24.1-1.el7.x86_64.net.core.rmem_default = 212992net.core.rmem_max = 268435456 log_fifo_size(268435456); source s_network {network(ip("0.0.0.0") port(514) transport("udp") so_rcvbuf(441326592) so-reuseport(1) persist-name("udp1"));network(ip("0.0.0.0") port(514) transport("udp") so_rcvbuf(441326592) so-reuseport(1) persist-name("udp2"));network(ip("0.0.0.0") port(514) transport("udp") so_rcvbuf(441326592) so-reuseport(1) persist-name("udp3"));network(ip("0.0.0.0") port(514) transport("udp") so_rcvbuf(441326592) so-reuseport(1) persist-name("udp4"));network(ip("0.0.0.0") port(514) transport("tcp") max_connections(200) keep_alive(yes) so_rcvbuf(67108864));}; We are limited to UDP, unfortunately, because we do not have control over the devices/networks/etc. that are sending to us, but we have changed as many of the internal senders and destinations to TCP as we can. With a script I created to view the packets, including drops, as well as the individual RECVQs, the issue can be illustrated. Here's what things look like normally:Thu May 7 10:48:15 EDT 2020, 27003 IP pkts rcvd,26980 IP pkts sent,24951 UDP pkts rcvd, 28075 UDP pkts sent,0 UDP pkt rcv err RECVQ-1=2176RECVQ-2=0RECVQ-3=0RECVQ-4=0Thu May 7 10:48:16 EDT 2020, 28453 IP pkts rcvd,28426 IP pkts sent,26185 UDP pkts rcvd, 29180 UDP pkts sent,0 UDP pkt rcv errRECVQ-1=0RECVQ-2=0RECVQ-3=4352RECVQ-4=0Thu May 7 10:48:17 EDT 2020, 28294 IP pkts rcvd,28276 IP pkts sent,26277 UDP pkts rcvd, 28709 UDP pkts sent,0 UDP pkt rcv errRECVQ-1=2176RECVQ-2=0RECVQ-3=0RECVQ-4=0 The RECVQs are sparsely used, and there are no errors. Around 9pm every night, the packet counts go up significantly (probably due to backup related logs):Wed May 6 21:00:08 EDT 2020, 66382 IP pkts rcvd,66366 IP pkts sent,39405 UDP pkts rcvd, 67592 UDP pkts sent,0 UDP pkt rcv errRECVQ-1=1595008RECVQ-2=106217088RECVQ-3=53694976RECVQ-4=31858816Wed May 6 21:00:09 EDT 2020, 69317 IP pkts rcvd,69338 IP pkts sent,44446 UDP pkts rcvd, 75958 UDP pkts sent,0 UDP pkt rcv errRECVQ-1=13056RECVQ-2=126397312RECVQ-3=75568128RECVQ-4=41626880Wed May 6 21:00:10 EDT 2020, 71205 IP pkts rcvd,71227 IP pkts sent,43657 UDP pkts rcvd, 74603 UDP pkts sent,0 UDP pkt rcv errRECVQ-1=920448RECVQ-2=146122752RECVQ-3=100951168RECVQ-4=52622208Wed May 6 21:00:12 EDT 2020, 69578 IP pkts rcvd,69454 IP pkts sent,124465 UDP pkts rcvd, 163367 UDP pkts sent,0 UDP pkt rcv errRECVQ-1=13140864RECVQ-2=44494848RECVQ-3=125579136RECVQ-4=0 Still, though, it's handling it with no errors. But then at some point a threshold is reached and errors start piling up:Wed May 6 21:00:20 EDT 2020, 63177 IP pkts rcvd,63291 IP pkts sent,0 UDP pkts rcvd, 0 UDP pkts sent,38011 UDP pkt rcv errRECVQ-1=536871424RECVQ-2=200357376RECVQ-3=292948352RECVQ-4=28890752Wed May 6 21:00:21 EDT 2020, 69501 IP pkts rcvd,69464 IP pkts sent,0 UDP pkts rcvd, 1 UDP pkts sent,42158 UDP pkt rcv errRECVQ-1=536871424RECVQ-2=223551360RECVQ-3=314995584RECVQ-4=41735680Wed May 6 21:00:23 EDT 2020, 69962 IP pkts rcvd,69978 IP pkts sent,0 UDP pkts rcvd, 2 UDP pkts sent,43775 UDP pkt rcv errRECVQ-1=536871424RECVQ-2=244732544RECVQ-3=338239616RECVQ-4=53858176Wed May 6 21:00:24 EDT 2020, 68266 IP pkts rcvd,68216 IP pkts sent,0 UDP pkts rcvd, 0 UDP pkts sent,43118 UDP pkt rcv errRECVQ-1=536871424RECVQ-2=265258752RECVQ-3=360643712RECVQ-4=65362688 The common denominator I've found is that one of the RECVQs hits 536,871,424. This number seems to be almost exactly double (512 difference) the rmem.max/log_fifo_size (268,435,456). Even though there seems to be capacity in the other RECVQs, just one of those hitting that magic number seems to be enough to throw things out of whack. During this time, the CPU usage for syslog-ng also drops: top - 21:00:11 up 2 days, 8:40, 2 users, load average: 1.09, 1.41, 1.32 PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 7214 root 20 0 2111388 1.7g 5392 S 123.6 10.9 3080:46 syslog-ng top - 21:00:14 up 2 days, 8:40, 2 users, load average: 1.00, 1.39, 1.31 PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 7214 root 20 0 2111388 1.7g 5392 S 24.6 10.9 3080:46 syslog-ng top - 21:00:17 up 2 days, 8:40, 2 users, load average: 1.00, 1.39, 1.31 PID USER PR NI VIRT RES SHR S %CPU %MEM TIME+ COMMAND 7214 root 20 0 2111388 1.7g 5392 S 7.6 10.9 3080:47 syslog-ng As best I can tell, based on the reading I've done, the message counts we're getting should be doable, but it's still not clear exactly how to size some of these options as they relate to UDP (like log_fetch_limit, which we do not have set). Anything else I should try?

2 4

parsing cisco firepower logs problem with 3.33
by Stoffel, John (TAI) 09 Mar '22

09 Mar '22

Hi, I'm trying to parse some cisco logs from a Cisco firepower firewall, using syslog-ng v3.33 on a CentOS 7 system. After pounding my head against the wall a few times to realize that you can't just re-start syslog-ng and have it re-read a source file from scratch... that instead I need to just push the data using netcat, it's now in a state where I think I can try to debug things. My logs look like this: <166>2022-02-16T15:31:53Z na-zy-int-fp1140-p02 : %FTD-6-305012: Teardown dynamic UDP translation fr om TAI-INSIDE:1.2.3.110/51288 to FOO-OUTSIDE:6.7.8.18/33333 duration 0:00:00 <166>2022-02-16T15:31:53Z na-zy-int-fp1140-p02 : %FTD-6-305012: Teardown dynamic TCP translation fr om FOO-WAN_IN:10.92.60.80/59877 to FOO-OUTSIDE:6.7.8.18/59877 duration 0:01:01 <166>2022-02-16T15:31:53Z na-zy-int-fp1140-p02 : %FTD-6-305011: Built dynamic UDP translation from FOO-INSIDE:1.2.3.110/51288 to FOO-OUTSIDE:6.7.8.18/5632 Looking at this log, vs the examples given in the /usr/share/syslog-ng/include/scl/cisco/plugin.conf file, I think the problem is that my logs shows the: sequence, date: origin, %MSG instead of sequence, origin, date: %MSG and it's not clear to me how I would hack the plugin.conf file to handle this issue. My end goal is to be able to parse the message enough by log level so I can forward only a subset of messages to another remote syslog system. Thanks, John Sr. Storage Architect TOSHIBA AMERICA, INC. 1251 6th, Ave 41st flr, New York, NY 10020 508-736-5499 (mobile) E-Mail: john.stoffel(a)toshiba.com<mailto:john.stoffel@toshiba.com> Website: Service Now Self Service Portal<https://nassc.service-now.com/ess/navpage.do>

2 7

Syslog-ng v3.31 destination option not working
by Maurya, Shivani 25 Feb '22

25 Feb '22

Hi, I am referring the syslog-ng admin guide - https://www.syslog-ng.com/technical-documents/doc/syslog-ng-open-source-edi… and using the snmp() destination option. The following is included in my syslog-ng.conf file as part of destination - destination d_snmpv2c{ snmp( version('v2c') host('192.168.1.1') trap-obj('.1.3.6.1.6.3.1.1.4.1.0', 'Objectid', '.1.3.6.1.4.1.18372.3.1.1.1.2.1') snmp-obj('.1.3.6.1.4.1.18372.3.1.1.1.1.1.0', 'Octetstring', 'Test SNMP trap') snmp-obj('.1.3.6.1.4.1.18372.3.1.1.1.1.2.0', 'Octetstring', 'admin') snmp-obj('.1.3.6.1.4.1.18372.3.1.1.1.1.3.0', 'Ipaddress', '192.168.1.1') ); }; But syslog-ng service/daemon is not starting & pointing error at snmp(). Can you please help on this? FYI - I am using RHEL7. Regards, Shivani Maurya

3 5

allowed concurrent connections - bug?
by Evan Rempel 18 Feb '22

18 Feb '22

I am having an issue that is a little difficult to reproduce so I wanted some input from others. I have a syslog-ng 3.35.1 that has a TLS source defined with max-connections(10000) After some time the server starts logging a lot of messages syslog-ng[12802]: Number of allowed concurrent connections reached, rejecting connection; client='AF_INET(XXXX:61062)', local='AF_INET(YYYY:6514)', group_name='client_network_tcp', location='/etc/syslog-ng/syslog-ng.server.conf:61:9', max='10000' To the best of my ability I can only find about 2500 actual connections. Both lsof and netstat report around the 2500 connections. I had to restart syslog-ng to stop this situation. Has anyone seen this behavior before? I get a lot of TLS connections without a certificate. Error reading RFC6587 style framed data Pperhaps the counters are not decremented for those timed out connections? -- Evan Rempel

2 5

Yocto: build syslog-ng with Python support
by Christian Hollermann 16 Feb '22

16 Feb '22

Hi, in our project we use a Yocto based build system (branch dunfell, Linux Kernel 5.10) for a while. At the next step, we plan to capture log messages from different sources with the help of syslog-ng (3.24.1). For some sources we will use a Python script that implements an interface to another system. Our problem is now, that we cannot build syslog-ng with a working Python support. Our build completes successfully, but the newly generated library 'libmod-python.so' seems to be broken as we get an error message when we use the following command to verify the installation. syslog-ng --module-registry | grep python Error opening plugin module; module='mod-python', error='/usr/lib/syslog-ng/libmod-python.so: undefined symbol: PyExc_ValueError' Module: mod-python Status: Unable to dlopen shared object, probably not a syslog-ng module Below you can see our current syslog recipe. syslog-ng_%.bbappend: FILESEXTRAPATHS_append := ":${THISDIR}/${PN}" DEPENDS += " python3 " # Override of default config file SRC_URI += "file://syslog-ng.conf.systemd" # Just to be sure EXTRA_OECONF_remove = "--disable-python" # Enable Python Support EXTRA_OECONF += "--enable-python --with-python=3" Can you give us hint, how to to fix this behavior? Thanks and Regards, Christian

1 0

syslog-ng OSE v3.30.1 - slowdown when using env variable in destination clause
by Wilson, Jonathan L 14 Feb '22

14 Feb '22

I have a pair of syslog-ng OSE servers on different LANs, which forward some of the log entries that they receive to separate collectors. In other words, Syslog-A -> collector-A Syslog-B -> collector-B I would like for the two syslog servers to use a common syslog-ng.conf for obvious reasons. I first built destination clauses with the separate IPs, like so: destination d_collector { udp( "192.168.1.23" port(514) template("${DATE} ${HOST} ${MSG}") ); }; This worked fine. But when I define an environment variable, COLLECTOR_IP, and then change my configuration to: destination d_collector { udp( "`COLLECTOR_IP`" port(514) template("${DATE} ${HOST} ${MSG}") ); }; Then syslog-ng's throughput slows down to a trickle (but does continue); CPU utilization (ordinarily substantial - this is a busy log handler) drops to near zero. Changing the configurations back to use the raw IP address restored previous throughput. There are no unusual messages in syslog-ng's own logfile. Anyone seen this puzzling behavior? Thanks, Jon Wilson jonathan.wilson(a)vumc.org<mailto:jonathan.wilson@vumc.org>

2 1

syslog-ng 3.35.1-2.el7
by Evan Rempel 10 Feb '22

10 Feb '22

The packaged syslog-ng RPM for RHEL7 from the stable copr repo https://copr-be.cloud.fedoraproject.org/results/czanik/syslog-ng-stable fails to have a dependency specified for glib library with a large enough version to satisfy the runtime requirements. The syslog-ng package has a requirement listed of "libglib-2.0.so.0()(64bit)" but if there is an older glib2 package installed that satisfies this requirement syslog-ng will fail to start with an error /usr/sbin/syslog-ng: symbol lookup error: /usr/lib64/libsyslog-ng-3.35.so.0: undefined symbol: g_ptr_array_find_with_equal_func Can the package maintainer add a >= version specification to the requirements so that the install of syslog-ng will automatically update to a version of glib2 that works with syslog-ng? -- Evan Rempel

4 5

Insider 2022-01: Reboot; Sequence; Monterey; CentOS 9;
by Peter Czanik (pczanik) 10 Feb '22

10 Feb '22

Dear syslog-ng users, This is the 98th issue of syslog-ng Insider, a monthly newsletter that brings you syslog-ng-related news. NEWS syslog-ng relaunch ------------------ Balázs Scheidler, founder of the syslog-ng project, started a new blog where he details why and how he started to work on syslog-ng even more actively. He also asks for your feedback! “syslog-ng has been around for decades: I started coding the first version of syslog-ng in September 1998, circa 24 years ago. The adoption of syslog-ng skyrocketed soon after that: people installed it in place of the traditional syslogd across the globe. It was packaged for Debian, Gentoo, SUSE and even commercial UNIXes. It became a default logging daemon in some of these Linux distributions. Commercial products started embedding it as a system component. Over the years however I feel that syslog-ng has become a trusted piece of infrastructure, few people really care about. I set out to change that.” Read the rest of the blog at https://syslog-ng-future.blog/syslog-ng-relaunch/ Sequence – making PatternDB creation for syslog-ng easier --------------------------------------------------------- We are well into the 21st century, but most of the log messages still arrive in an unstructured format. For well over a decade, syslog-ng had a solution to turn unstructured messages into name-value pairs, called PatternDB. However, creating a pattern database for PatternDB from scratch is a source of major pain. Or rather, it was: sequence-rtg – a fork of the sequence log analyzer – provides a new hope! It can easily create ready-to-use patterns for your most frequent log messages. Sequence-rtg is still in beta phase, and therefore is a bit rough around the edges. However, once you deal with the initial struggles of creating the database, it works just fine. Especially if you have lots of log messages. My experience was that the more log messages and larger batch sizes I had, the better quality patterns were generated. https://www.syslog-ng.com/community/b/blog/posts/sequence-making-patterndb-… Syslog-ng on MacOS Monterey --------------------------- Each new MacOS release brings some surprises when it comes to compiling syslog-ng. Just a couple of months ago, I provided you with a couple of pointers on how to compile syslog-ng on MacOS. Since then, MacOS Monterey was released and Homebrew was updated. So, here are some updated instructions for MacOS Monterey. https://www.syslog-ng.com/community/b/blog/posts/syslog-ng-on-macos-monterey Installing syslog-ng on CentOS Stream 9 --------------------------------------- CentOS Stream 9 has been around for a while, but it was officially announced just a few days ago. I already tested some earlier snapshots and they had some rough edges. The current version installed without random crashes, has networking and runs smoothly. EPEL – the semi-official repository by Fedora maintainers – is already there, but practically empty, syslog-ng or it’s dependencies are not yet there. As someone asked about syslog-ng support, I had a first try at building it. https://www.syslog-ng.com/community/b/blog/posts/installing-syslog-ng-on-ce… WEBINARS * You can browse recordings of past webinars at https://www.syslog-ng.com/events/ Your feedback and news, or tips about the next issue are welcome. To read this newsletter online, visit: https://syslog-ng.com/blog/ Peter Czanik (CzP) <peter.czanik(a)oneidentity.com> Balabit (a OneIdentity company) / syslog-ng upstream https://syslog-ng.com/community/ https://twitter.com/PCzanik

1 0