fluent-plugin-grok-parser
fluent-plugin-grok-parser copied to clipboard
Syslog TAGs exceeding 32 character will not grok
Hello Everyone! I would like some feedback on an issue i've observed with syslog tags exceeding 32 characters. The syslog RFC defines a "TAG" in the syslog message (typically program name and pid) must not exceed 32 characters. Rsyslog will truncate any part of the TAG past 32 characters when shipping logs (at least on my systems Cent/Alma). Resulting in something like this:
original message:
<77>Aug 18 05:00:03 my-test-host run-parts(/etc/cron.daily)[25833]: finished prelink
what fluentd sees:
<77>Aug 18 05:00:03 my-test-host run-parts(/etc/cron.daily)[25833 finished prelink
You'll notice that run-parts(/etc/cron.daily)[25833
is exactly 32 characters. I suggest a change to SYSLOGBASE that accounts for this in case rsyslog or any other shipping mechanism enforces the rfc standard. Something along the lines of:
SYSLOG_LONGTAG [\x21-\x5a\x5c\x5e-\x7e\x5b\x5d]{32}
SYSLOGPROG (%{PROG:[process][name]}(?:\[%{POSINT:[process][pid]:integer}\])?|%{SYSLOG_LONGTAG:syslog_longtag})
Thoughts? Kudos to @tlaberge-godaddy for finding this.