Using syslog-ng to forward logs is pretty nice, there’s plenty of documentation and the configuration is relatively easy to understand compared to other stuff out there (looking at you rsyslog), but that doesn’t mean everything is completely obvious. If you search for information on how to read a text file log with syslog-ng, you might come up with something like this:

source s_squid3access {
    file("/var/log/squid3/access.log" follow-freq(1)); };

Which checks the file /var/log/squid3/access.log for new entries every second. However, if you simply send this as is, you might end up with a message similar to the one below being sent to your syslog destination (note that I’ve modified my squid instance to log in the Apache Combined log format)

<13>1 2017-11-07T19:07:44+00:00 myproxy 192.168.1.4 - - [meta sequenceId="84"] - - [07/Nov/2017:19:07:43 +0000] "CONNECT www.netflix.com:443 HTTP/1.1" 200 12237 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/62.0.3202.75 Safari/537.36" TCP_MISS:HIER_DIRECT

which corresponds to the following line in the log file:

192.168.1.4 - - [07/Nov/2017:19:09:03 +0000] "CONNECT www.netflix.com:443 HTTP/1.1" 200 5101 "-" "Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/62.0.3202.75 Safari/537.36" TCP_MISS:HIER_DIRECT

Note the position of the IP address in the syslog message – it is in the syslog header section in the position of the syslog program field. If you want to collect and parse this information in a SIEM for example, this will cause you quite the headache, as the message it attempts to interpret will begin “– – [07/Nov/2017:19:07:43 +0000]” rather than “192.168.1.4 – – [07/Nov/2017:19:07:43 +0000]“. It will no longer match the format of the data the collector is expecting (probably resulting in the message not getting parsed at all), and even if you tried fixing that, the missing data is a fairly crucial bit of info – there’s not much use knowing that someone visited a site if you can’t find out which user or host it was.

The answer is given in the syslog-ng documentation, although it is not immediately obvious. The section on collecting messages from text files hints that if the message does not have a syslog header, it may behave in an unusual way, but it does not explain in detail what will happen; for that you must look at the options for the file() method. In the description of the flags() option “no-parse” it notes that by default, syslog-ng assumes the first value in a log line is the syslog program. If you set this flag, your originating IP will again be part of the message section, and your SIEM/parsing will be happy again.

You can also set the program_override() option so that the program field is populated, as it is useful in certian SIEM/collection tools to have this info. Now your config file might look a bit like this:

source s_squid3access {
    file("/var/log/squid3/access.log" follow-freq(1) flags(no-parse) program_override("squid3")); };

and all should be well. Happy logging!