In syslog-ng OSE version 3.8 and later, you can use an external database file to add additional metadata to your log messages. For example, you can create a database (or export it from an existing tool) that contains a list of hostnames or IP addresses, and the department of your organization that the host belongs to, the role of the host (mailserver, webserver, and so on), or similar contextual information.
The database file is a simple text file in comma-separated value (CSV) format, where each line contains the following information:
-
A selector or ID that appears in the log messages, for example, the hostname. To use shell-style globbing (wildcards) in selectors, see Shell-style globbing in the selector. You can also reference the name of a filter that matches the messages, see Using filters as selector
-
The name of the name-value pair that syslog-ng OSE adds to matching log messages.
-
The value of the name-value pairs. Starting with syslog-ng OSE version 3.22, the value of the name-value pair can be a template or a template function, for example, "selector3,name,$(echo $HOST_FROM)";
For example, the following csv-file contains three lines identified with the IP address, and adds the host-role field to the log message.
192.168.1.1,host-role,webserver
192.168.2.1,host-role,firewall
192.168.3.1,host-role,mailserver
The database file:
The database file must comply with the RFC4180 CSV format, with the following exceptions and limitations:
To add multiple name-value pairs to a message, include a separate line in the database for each name-value pair, for example:
192.168.1.1,host-role,webserver
192.168.1.1,contact-person,"John Doe"
192.168.1.1,contact-email,johndoe@example.com
Technically, add-contextual-data() is a parser in syslog-ng OSE so you have to define it as a parser object.
Declaration:
parser p_add_context_data {
add-contextual-data(
selector("${HOST}"),
database("context-info-db.csv"),
);
};
You can also add data to messages that do not have a matching selector entry in the database using the default-selector() option.
If you modify the database file, you have to reload syslog-ng OSE for the changes to take effect. If reloading syslog-ng OSE or the database file fails for some reason, syslog-ng OSE will keep using the last working database file.
Example: Adding metadata from a CSV file
The following example defines uses a CSV database to add the role of the host based on its IP address, and prefixes the added name-value pairs with .metadata. The destination includes a template that simply appends the added name-value pairs to the end of the log message.
@include "scl.conf"
source s_network {
network(port(5555));
};
destination d_local {
file("/tmp/test-msgs.log"
template("$MSG Additional metadata:[${.metadata.host-role}]")};
parser p_add_context_data {
add-contextual-data(selector("$SOURCEIP"), database("context-info-db.csv"), default-selector("unknown"), prefix(".metadata."));
};
log {
source(s_network);
parser(p_add_context_data);
destination(d_local);
};192.168.1.1,host-role,webserver
192.168.2.1,host-role,firewall
192.168.3.1,host-role,mailserver
unknown,host-role,unknown
To better control to which log messages you add contextual data, you can use filters as selectors. In this case, the first column of the CSV database file must contain the name of a filter. For each message, syslog-ng OSE evaluates the filters in the order they appear in the database file. If a filter matches the message, syslog-ng OSE adds the name-value pair related to the filter.
For example, the database file can contain the entries. (For details on the accepted CSV-format, see database().)
f_auth,domain,all
f_localhost,source,localhost
f_kern,domain,kernel
Note that syslog-ng OSE does not evaluate other filters after the first match. For example, if you use the previous database file, and a message matches both the f_auth and f_localhost filters, syslog-ng OSE adds only the name-value pair of f_auth to the message.
To add multiple name-value pairs to a message, include a separate line in the database for each name-value pair, for example:
f_localhost,host-role,firewall
f_localhost,contact-person,"John Doe"
f_localhost,contact-email,johndoe@example.com
You can also add data to messages that do not have a matching selector entry in the database using the default-selector() option.
You must store the filters you reference in a database in a separate file. This file is similar to a syslog-ng OSE configuration file, but must contain only a version string and filters (and optionally comments). You can use the syslog-ng --syntax-only <filename> command to ensure that the file is valid. For example, the content of such a file can be:
@version: 3.38
filter f_localhost { host("mymachine.example.com") };
filter f_auth { facility(4) };
filter f_kern { facility(0) };
Declaration:
parser p_add_context_data_filter {
add-contextual-data(
selector(filters("filters.conf")),
database("context-info-db.csv"),
prefix(".metadata.")
);
};
If you modify the database file, or the file that contains the filters, you have to reload syslog-ng OSE for the changes to take effect. If reloading syslog-ng OSE or the files fails for some reason, syslog-ng OSE will keep using the last working version of the file.
Starting with in syslog-ng OSE3.24 and later, you can use shell-style globbing ('*' and '?' wildcards) in the selector.
To use globs in a selector
-
Use the glob() option within the selector() option in your syslog-ng OSE configuration file, for example:
parser p_add_context_data {
add-contextual-data(
selector(glob("${HOST}"))
database("context-info-db.csv")
);
};
-
Use globs and wildcards in the selector column of your CSV-file, for example:
example-glob-entry1*,sourcetype,:hec:user
example-glob-entry2*,sourcetype,:hec:user
postfix*,sourcetype,:hec:mta
Note the following points when using globbing in the selector:
-
The order of the patterns depends on the CSV-file. The order of entries in the database determines the matching order.
-
The globs are matched against the expanded template string sequentially.
-
Put more specific patterns to the top of the CSV-file. The syslog-ng OSE appication does not evaluate other entries after the first match.
-
In debug mode, syslog-ng OSE sends log messages to its internal() destination to help troubleshooting. For example:
[2019-09-21T06:01:10.748237] add-contextual-data(): Evaluating glob against message; glob-template='$PROGRAM', string='postfix/smtpd', pattern='example-glob-entry1*', matched='0'
[2019-09-21T06:01:10.748562] add-contextual-data(): Evaluating glob against message; glob-template='$PROGRAM', string='postfix/smtpd', pattern='example-glob-entry2*', matched='0'
[2019-09-21T06:01:10.748697] add-contextual-data(): Evaluating glob against message; glob-template='$PROGRAM', string='postfix/smtpd', pattern='postfix*', matched='1'
[2019-09-21T06:01:10.750084] add-contextual-data(): message lookup finished; message='almafa', resolved_selector='postfix*', selector='postfix*', msg='0x8e15320'
The add-contextual-data() has the following options.
Required options:
The following options are required: selector(), database().
database()
| Type: |
<path-to-file>.csv |
| Default: |
|
Description: Specifies the path to the CSV file, for example, /opt/syslog-ng/my-csv-database.csv. The extension of the file must be .csv, and can include Windows-style (CRLF) or UNIX-style (LF) linebreaks. You can use absolute path, or relative to the syslog-ng OSE binary.
default-selector()
| Synopsis: |
default-selector() |
Description: Specifies the ID of the entry (line) that is corresponds to log messages that do not have a selector that matches an entry in the database. For example, if you add name-value pairs from the database based on the hostname from the log message (selector("${HOST}")), then you can include a line for unknown hosts in the database, and set default-selector() to the ID of the line for unknown hosts. In the CSV file:
unknown-hostname,host-role,unknown
In the syslog-ng OSE configuration file:
add-contextual-data(
selector("$HOST")
database("context-info-db.csv")
default-selector("unknown-hostname")
);
ignore-case()
| Synopsis: |
ignore-case() |
|
Default: |
ignore-case(no) |
Description: Specifies if selectors are handled as case insensitive. If you set the ignore-case() option to yes, selectors are handled as case insensitive.
prefix()
Description: Insert a prefix before the name part of the added name-value pairs (including the pairs added by the default-selector()) to help further processing.
selector()
Description: Specifies the string or macro that syslog-ng OSE evaluates for each message, and if its value matches the ID of an entry in the database, syslog-ng OSE adds the name-value pair of every matching database entry to the log message. You can use the following in the selector() option.
