MIT License

A parser for Syslog messages and transports.

Blazing fast Syslog parsers

By @leodido.

To wrap up, this package provides:

This library provides the pieces to parse Syslog messages transported following various RFCs.

For example:

  • TLS with octet count (RFC5425)
  • TCP with non-transparent framing or with octet count (RFC 6587)
  • UDP carrying one message per packet (RFC5426)

Installation

go get github.com/influxdata/go-syslog/v3

Docs

Documentation

The docs directory contains .dot files representing the finite-state machines (FSMs) implementing the syslog parsers and transports.

Usage

Suppose you want to parse a given sequence of bytes as a RFC5424 message.

Notice that the same interface applies for RFC3164. But you can always take a look at the examples file.

i := []byte(`<165>4 2018-10-11T22:14:15.003Z mymach.it e - 1 [[email protected] iut="3"] An application event log entry...`)
p := rfc5424.NewParser()
m, e := p.Parse(i)

This results in m being equal to:

// (*rfc5424.SyslogMessage)({
//  Base: (syslog.Base) {
//   Facility: (*uint8)(20),
//   Severity: (*uint8)(5),
//   Priority: (*uint8)(165),
//   Timestamp: (*time.Time)(2018-10-11 22:14:15.003 +0000 UTC),
//   Hostname: (*string)((len=9) "mymach.it"),
//   Appname: (*string)((len=1) "e"),
//   ProcID: (*string)(<nil>),
//   MsgID: (*string)((len=1) "1"),
//   Message: (*string)((len=33) "An application event log entry...")
//  },
//  Version: (uint16) 4,
//  StructuredData: (*map[string]map[string]string)((len=1) {
//   (string) (len=8) "[email protected]": (map[string]string) (len=1) {
//    (string) (len=3) "iut": (string) (len=1) "3"
//   }
//  })
// })

And e being equal to nil since the i byte slice contains a perfectly valid RFC5424 message.

Best effort mode

RFC5424 parser has the ability to perform partial matches (until it can).

With this mode enabled, when the parsing process errors out it returns the message collected until that position, and the error that caused the parser to stop.

Notice that in this modality the output is returned iff it represents a minimally valid message – ie., a message containing almost a priority field in [1,191] within angular brackets, followed by a version in ]0,999] (in the case of RFC5424).

Let’s look at an example.

i := []byte("<1>1 A - - - - - -")
p := NewParser(WithBestEffort())
m, e := p.Parse(i)

This results in m being equal to the following SyslogMessage instance.

// (*rfc5424.SyslogMessage)({
//  Base: (syslog.Base) {
//   Facility: (*uint8)(0),
//   Severity: (*uint8)(1),
//   Priority: (*uint8)(1),
//   Timestamp: (*time.Time)(<nil>),
//   Hostname: (*string)(<nil>),
//   Appname: (*string)(<nil>),
//   ProcID: (*string)(<nil>),
//   MsgID: (*string)(<nil>),
//   Message: (*string)(<nil>)
//  },
//  Version: (uint16) 1,
//  StructuredData: (*map[string]map[string]string)(<nil>)
// })

And, at the same time, in e reporting the error that actually stopped the parser.

// expecting a RFC3339MICRO timestamp or a nil value [col 5]

Both m and e have a value since at the column the parser stopped it already was able to construct a minimally valid RFC5424 SyslogMessage.

Builder

This library also provides a builder to construct valid syslog messages.

Notice that its API ignores input values that does not match the grammar.

Let’s have a look to an example.

msg := &rfc5424.SyslogMessage{}
msg.SetTimestamp("not a RFC3339MICRO timestamp")
msg.Valid() // Not yet a valid message (try msg.Valid())
msg.SetPriority(191)
msg.SetVersion(1)
msg.Valid() // Now it is minimally valid

Printing msg you will verify it contains a nil timestamp (since an invalid one has been given).

// (*rfc5424.SyslogMessage)({
//  Base: (syslog.Base) {
//   Facility: (*uint8)(23),
//   Severity: (*uint8)(7),
//   Priority: (*uint8)(191),
//   Timestamp: (*time.Time)(<nil>),
//   Hostname: (*string)(<nil>),
//   Appname: (*string)(<nil>),
//   ProcID: (*string)(<nil>),
//   MsgID: (*string)(<nil>),
//   Message: (*string)(<nil>)
//  },
//  Version: (uint16) 1,
//  StructuredData: (*map[string]map[string]string)(<nil>)
// })

Finally you can serialize the message into a string.

str, _ := msg.String()
// <191>1 - - - - - -

Message transfer

Excluding encapsulating one message for packet in packet protocols there are two ways to transfer syslog messages over streams.

The older – ie., the non-transparent framing – and the newer one – ie., the octet counting framing – which is reliable and has not been seen to cause problems noted with the non-transparent one.

This library provide stream parsers for both.

Octet counting

In short, RFC5425 and RFC6587, aside from the protocol considerations, describe a transparent framing technique for Syslog messages that uses the octect counting technique – ie., the message length of the incoming message.

Each Syslog message is sent with a prefix representing the number of bytes it is made of.

The octecounting package parses messages stream following such rule.

To quickly understand how to use it please have a look at the example file.

Non transparent

The RFC6587 also describes the non-transparent framing transport of syslog messages.

In such case the messages are separated by a trailer, usually a line feed.

The nontransparent package parses message stream following such technique.

To quickly understand how to use it please have a look at the example file.

Things we do not support:

  • trailers other than LF or NUL
  • trailers which length is greater than 1 byte
  • trailer change on a frame-by-frame basis

Performances

To run the benchmark execute the following command.

make bench

On my machine1 these are the results obtained paring RFC5424 syslog messages with best effort mode on.

[no]_empty_input__________________________________  4524100        274 ns/op      272 B/op        4 allocs/op
[no]_multiple_syslog_messages_on_multiple_lines___  3039513        361 ns/op      288 B/op        8 allocs/op
[no]_impossible_timestamp_________________________  1244562        951 ns/op      512 B/op       11 allocs/op
[no]_malformed_structured_data____________________  2389249        512 ns/op      512 B/op        9 allocs/op
[no]_with_duplicated_structured_data_id___________  1000000       1183 ns/op      712 B/op       17 allocs/op
[ok]_minimal______________________________________  6876235        178 ns/op      227 B/op        5 allocs/op
[ok]_average_message______________________________   730473       1653 ns/op     1520 B/op       24 allocs/op
[ok]_complicated_message__________________________   908776       1344 ns/op     1264 B/op       24 allocs/op
[ok]_very_long_message____________________________   392737       3114 ns/op     2448 B/op       25 allocs/op
[ok]_all_max_length_and_complete__________________   510740       2431 ns/op     1872 B/op       28 allocs/op
[ok]_all_max_length_except_structured_data_and_mes   755124       1593 ns/op      867 B/op       13 allocs/op
[ok]_minimal_with_message_containing_newline______  6142984        199 ns/op      230 B/op        6 allocs/op
[ok]_w/o_procid,_w/o_structured_data,_with_message  1670286        732 ns/op      348 B/op       10 allocs/op
[ok]_minimal_with_UTF-8_message___________________  3013480        407 ns/op      339 B/op        6 allocs/op
[ok]_minimal_with_UTF-8_message_starting_with_BOM_  2926410        423 ns/op      355 B/op        6 allocs/op
[ok]_with_structured_data_id,_w/o_structured_data_  1558971        814 ns/op      570 B/op       11 allocs/op
[ok]_with_multiple_structured_data________________  1000000       1243 ns/op     1205 B/op       16 allocs/op
[ok]_with_escaped_backslash_within_structured_data  1000000       1025 ns/op      896 B/op       17 allocs/op
[ok]_with_UTF-8_structured_data_param_value,_with_  1000000       1241 ns/op     1034 B/op       19 allocs/op

As you can see it takes:

  • ~250ns to parse the smallest legal message

  • less than 2µs to parse an average legal message

  • ~3µs to parse a very long legal message

Other RFC5424 implementations, like this one in Rust, spend 8µs to parse an average legal message.

TBD: comparison against other Go parsers.


  • [1]: Intel Core i7-8850H CPU @ 2.60GHz

GitHub

https://github.com/influxdata/go-syslog