Heroku Logstash config

logstash.conf

input {
  # Testing:
  # tcp {
  #   type => "heroku"
  #   port => 3333
  # }
  # stdin { type => "stdin"}

  tcp {
    type => "heroku"
    port => 1514
  }
}

filter {
  if [type] == "heroku" {
  
    grok {
      match => {"message" => "^%{POSINT:syslog_octet_size:int} %{SYSLOG5424PRI}%{NONNEGINT:syslog5424_ver} +(?:%{TIMESTAMP_ISO8601:syslog5424_ts}|-) +(?:%{HOSTNAME:syslog5424_host}|-) +(?:%{WORD:syslog5424_app}|-) +(?:%{HOSTNAME:syslog5424_proc}|-) +(?:%{WORD:syslog5424_msgid}|-) +(?:%{SYSLOG5424SD:syslog5424_sd}|-|) +%{GREEDYDATA:syslog5424_msg}" }
      add_field => [ "drain_token", "%{syslog5424_host}" ]
      add_field => [ "component", "%{syslog5424_app}" ]
      add_field => [ "received_at", "%{@timestamp}" ]
      add_field => [ "received_from", "%{host}" ]
    }
    grok {
      match => { "syslog5424_proc" => "%{WORD:process}(?:\.%{INT:instance:int})?" }
    }

    if !("_grokparsefailure" in [tags]) {
      date {
        match => [ "syslog5424_ts", "ISO8601" ]
      }
  
      multiline {
        # max syslog packet size
        pattern => "^1116 "
        stream_identity => "%{host}.%{@type}.%{component}.%{process}.%{instance}."
        what => next
      }
  
      if "multiline" in [tags] {
        mutate {
          join => ["syslog5424_msg", ""]
        }
      }

      mutate {
        replace => [ "message", "%{syslog5424_msg}" ]
      }

      if [component] == "heroku" {
        kv { add_tag => ["kv", "parsed"] }
      }

      if [message] =~ "^@cee" {
          mutate {
            add_tag => "cee"
            gsub => ["message", "^@cee:\s*", ""]
          }

          json {
            source => "message"
            target => "msg_data"
            add_tag => ["json", "parsed"]
          }
      }

      if "parsed" in [tags] {
        mutate {
          remove_field => ["message"]
        }
      }
    }
    else { 
      # not grok'd!
      mutate {
        add_field => [ "drain_token", "unknown" ]
        add_field => [ "component", "unknown" ]
      }
    }
  }
}
  
output { 
  # Testing:
  # stdout {
  #   debug => true
  #   debug_format => "json"
  # }

  elasticsearch { embedded => true }
  file {
    path => "./output-logs/%{drain_token}/%{component}/%{+YYYY-MM-dd}/logstash.log"
  }
}

Had an issue with SYSLOG5424LINE: it expects the app to be a WORD, when it could be WORD.INT (which is matched later, so maybe SYSLOG5424LINE was redefined earlier?)

Heroku's syslog may break up log lines into multiple syslog packets. The heuristic used above is that when a syslog packet is exactly 1116 (the max length used by Heroku), assume it will carry over to the next line. It is possible this will cause a problem with lines exactly that length.

JSON lines are indicated with the @Cee prefix (see http://cee.mitre.org/language/1.0-beta1/cls.html), which can be produced with Lograge::Formatters::Cee (https://github.com/roidrage/lograge)

Hi Jim. I am trying to set this up on a cluster set up at Elastic Cloud. What URL should I use for the log drain? Do you think this config still works for Heroku?