Log Formats#

Built-in Formats#

Log files loaded into lnav are parsed based on formats defined in configuration files. Many formats are already built in to the lnav binary and you can define your own using a JSON file. When loading files, each format is checked to see if it can parse the first few lines in the file. Once a match is found, that format will be considered that files format and used to parse the remaining lines in the file. If no match is found, the file is considered to be plain text and can be viewed in the “text” view that is accessed with the t key.

The following log formats are built into lnav:

Name

Table Name

Description

Common Access Log

access_log

The default web access log format for servers like Apache.

Amazon ALB log

alb_log

Log format for Amazon Application Load Balancers

Generic Block

block_log

A generic format for logs, like cron, that have a date at the start of a block.

Bunyan log

bunyan_log

Bunyan JSON logging library for node.js

Candlepin log format

candlepin_log

Log format used by Candlepin registration system

Yum choose_repo Log

choose_repo_log

The log format for the yum choose_repo tool.

Cloudflare Access Log

cloudflare_json_log

Cloudflare Enterprise detailed logs of metadata

CloudVM Ram Log

cloudvm_ram_log

Periodic dumps of ram sizes

CUPS log format

cups_log

Log format used by the Common Unix Printing System

Dpkg Log

dpkg_log

The debian dpkg log.

Amazon ELB log

elb_log

Log format for Amazon Elastic Load Balancers

engine log

engine_log

The log format for the engine.log files from RHEV/oVirt

Common Error Log

error_log

The default web error log format for servers like Apache.

ESXi Syslog

esx_syslog_log

Format specific to the ESXi syslog

Fsck_hfs Log

fsck_hfs_log

Log for the fsck_hfs tool on Mac OS X.

GitHub Events Log

github_events_log

Format for the public GitHub timeline from gharchive.org

Glog

glog_log

The google glog format.

HAProxy HTTP Log Format

haproxy_log

The HAProxy log format

Java log format

java_log

Log format used by log4j and output by most java programs

journalctl JSON log format

journald_json_log

Logger format as created by systemd journalctl -o json

Katello log format

katello_log

Log format used by katello and foreman as used in Satellite 6.

Nextcloud server logs

nextcloud

Nextcloud JSON server logs audit.log, flow.log, and nextcloud.log

OpenAM Log

openam_log

The OpenAM identity provider.

OpenAM Debug Log

openamdb_log

Debug logs for the OpenAM identity provider.

OpenStack log format

openstack_log

The log format for the OpenStack log files

CUPS Page Log

page_log

The CUPS server log of printed pages.

Papertrail Service

papertrail_log

Log format for the papertrail log management service

Packet Capture

pcap_log

Internal format for pcap files

Process State

procstate_log

Periodic dumps of process state

Redis

redis_log

The Redis database

S3 Access Log

s3_log

S3 server access log format

SnapLogic Server Log

snaplogic_log

The SnapLogic server log format.

SSSD log format

sssd_log

Log format used by the System Security Services Daemon

Strace

strace_log

The strace output format.

sudo

sudo_log

The sudo privilege management tool.

Syslog

syslog_log

The system logger format found on most posix systems.

TCF Log

tcf_log

Target Communication Framework log

TCSH History

tcsh_history

The tcsh history file format.

UniFi iptables log

unifi_iptables_log

The UniFi gateway iptables logger format (for /var/log/iptables).

UniFi log

unifi_log

The UniFi gateway messages logger format (for /var/log/messages).

Uwsgi Log

uwsgi_log

The uwsgi log format.

Vdsm Logs

vdsm_log

Vdsm log format

VMKernel Logs

vmk_log

The VMKernel’s log format

VMware Logs

vmw_log

One of the log formats used in VMware’s ESXi and vCenter software.

VMware vSphere log format

vmw_py_log

The log format for some VMware vSphere services

VMware Go Log

vmw_vc_svc_log

Log files for go-based logs

VMWare PostgreSQL

vpostgres_log

Format for vpostgresql log files with format ‘%m %c %x %d %u %r %p %l’

RHN server XMLRPC log format

xmlrpc_log

Generated by Satellite’s XMLRPC component

XSV Formats#

In addition to the above formats, the following self-describing formats are supported:

  • The Bro Network Security Monitor TSV log format is supported in lnav versions v0.8.3+. The Bro log format is self-describing, so lnav will read the header to determine the shape of the file.

  • The W3C Extended Log File Format is supported in lnav versions v0.10.0+. The W3C log format is self-describing, so lnav will read the header to determine the shape of the file.

JSON-lines#

Logs encoded as JSON-lines can be parsed and pretty-printed in lnav by creating a log format file. The format file is a bit simpler to create since it doesn’t require a regular expression to match plain text. Instead, the format defines the relevant fields and provides a line-format array that specifies how the fields in the JSON object should be displayed.

See the following formats that are built into lnav as examples:

logfmt#

There is also basic support for the logfmt convention for formatting log messages. Files that use this format must have the entire line be key/value pairs and the timestamp contained in a field named time or ts. If the file you’re using does not quite follow this formatting, but wraps logfmt data with another recognized format, you can use the logfmt2json(str) SQL function to convert the data into JSON for further analysis.

Defining a New Format#

New log formats can be defined by placing JSON configuration files in subdirectories of the /etc/lnav/formats and ~/.lnav/formats/ directories. The directories and files can be named anything you like, but the files must have the ‘.json’ suffix. A sample file containing the builtin configuration will be written to this directory when lnav starts up. You can consult that file when writing your own formats or if you need to modify existing ones. Format directories can also contain ‘.sql’ and ‘.lnav’ script files that can be used automate log file analysis.

Creating a Format Using Regex101.com (v0.11.0+)#

For plain-text log files, the easiest way to create a log format definition is to create the regular expression that recognizes log messages using https://regex101.com . Simply copy a log line into the test string input box on the site and then start editing the regular expression. When building the regular expression, you’ll want to use named captures for the structured parts of the log message. Any raw message text should be matched by a captured named “body”. Once you have a regex that matches the whole log message, you can use lnav’s “management CLI” to create a skeleton format file. The skeleton will be populated with the regular expression from the site and the test string, along with any unit tests, will be added to the “samples” list. The “regex101 import” management command is used to create the skeleton and has the following form:

lnav -m regex101 import <regex101-url> <format-name> [<regex-name>]

If the import was successful, the path to the new format file should be printed out. The skeleton will most likely need some changes to make it fully functional. For example, the kind properties for captured values default to string, but you’ll want to change them to the appropriate type.

Format File Reference#

An lnav format file must contain a single JSON object, preferably with a $schema property that refers to the format-v1.schema, like so:

{
    "$schema": "https://lnav.org/schemas/format-v1.schema.json"
}

Each format to be defined in the file should be a separate field in the top-level object. The field name should be the symbolic name of the format and consist only of alphanumeric characters and underscores. This value will also be used as the SQL table name for the log. The value for each field should be another object with the following fields:

title:

The short and human-readable name for the format.

description:

A longer description of the format.

url:

A URL to the definition of the format.

file-pattern:

A regular expression used to match log file paths. Typically, every file format will be tried during the detection process. This field can be used to limit which files a format is applied to in case there is a potential for conflicts.

regex:

This object contains sub-objects that describe the message formats to match in a plain-text log file. Each regex MUST only match one type of log message. It must not match log messages that are matched by other regexes in this format. This uniqueness requirement is necessary because lnav will “lock-on” to a regex and use it to match against the next line in a file. So, if the regexes do not uniquely match each type of log message, messages can be matched by the wrong regex. The “lock-on” behavior is needed to avoid the performance hit of having to try too many different regexes.

Note

Log files that contain JSON messages should not specify this field.

pattern:

The regular expression that should be used to match log messages. The PCRE2 library is used by lnav to do all regular expression matching.

module-format:

If true, this regex will only be used to parse message bodies for formats that can act as containers, such as syslog. Default: false.

json:

True if each log line is JSON-encoded.

converter:

An object that describes how an input file can be detected and then converted to a form that can be interpreted by lnav. For example, a PCAP file is in a binary format that cannot be handled natively by lnav. However, a PCAP file can be converted by tshark into JSON-lines that can be handled by lnav. So, this configuration describes how the input file format can be detected and converted. See Automatic File Conversion for more information.

header:

An object that describes how to match the header of the input file.

expr:

An object that contains SQLite expressions that can be used to check if the input file’s header is of this type. The property name is the name of the expression and the value is the expression. The expression is evaluated with the following variables:

:header:

The hex-encoded version of the header content.

:filepath:

The path to the input file.

size:

The minimum size of header that is needed to do the match.

command:

The command to execute to convert the input file.

line-format:

An array that specifies the text format for JSON-encoded log messages. Log files that are JSON-encoded will have each message converted from the raw JSON encoding into this format. Each element is either an object that defines which fields should be inserted into the final message string and or a string constant that should be inserted. For example, the following configuration will transform each log message object into a string that contains the timestamp, followed by a space, and then the message body:

[ { "field": "ts" }, " ", { "field": "msg" } ]

Note

Line-feeds at the end of a value are automatically stripped.

field:

The name or JSON-Pointer of the message field that should be inserted at this point in the message. The special __timestamp__ field name can be used to insert a human-readable timestamp. The __level__ field can be used to insert the level name as defined by lnav.

Tip

Use a JSON-Pointer to reference nested fields. For example, to include a “procname” property that is nested in a “details” object, you would write the field reference as /details/procname.

min-width:

The minimum width for the field. If the value for the field in a given log message is shorter, padding will be added as needed to meet the minimum-width requirement. (v0.8.2+)

max-width:

The maximum width for the field. If the value for the field in a given log message is longer, the overflow algorithm will be applied to try and shorten the field. (v0.8.2+)

auto-width:

Flag that indicates that the width of the field should automatically be set to the widest value seen. (v0.11.2)

align:

Specifies the alignment for the field, either “left” or “right”. If “left”, padding to meet the minimum-width will be added on the right. If “right”, padding will be added on the left. (v0.8.2+)

overflow:

The algorithm used to shorten a field that is longer than “max-width”. The following algorithms are supported:

abbrev:

Removes all but the first letter in dotted text. For example, “com.example.foo” would be shortened to “c.e.foo”.

truncate:

Truncates any text past the maximum width.

dot-dot:

Cuts out the middle of the text and replaces it with two dots (i.e. ‘..’).

(v0.8.2+)

timestamp-format:

The timestamp format to use when displaying the time for this log message. (v0.8.2+)

default-value:

The default value to use if the field could not be found in the current log message. The built-in default is “-“.

text-transform:

Transform the text in the field. Supported options are: none, uppercase, lowercase, capitalize

prefix:

Text to prepend to the value. If the value is empty, this prefix will not be added.

suffix:

Text to append to the value. If the value is empty, this suffix will not be added.

timestamp-field:

The name of the field that contains the log message timestamp. Defaults to “timestamp”.

timestamp-format:

An array of timestamp formats using a subset of the strftime conversion specification. The following conversions are supported: %a, %b, %L, %M, %H, %I, %d, %e, %k, %l, %m, %p, %y, %Y, %S, %s, %Z, %z. In addition, you can also use the following:

%L:

Milliseconds as a decimal number (range 000 to 999).

%f:

Microseconds as a decimal number (range 000000 to 999999).

%N:

Nanoseconds as a decimal number (range 000000000 to 999999999).

%q:

Seconds from the epoch as a hexidecimal number.

%i:

Milliseconds from the epoch.

%6:

Microseconds from the epoch.

timestamp-divisor:

For JSON logs with numeric timestamps, this value is used to divide the timestamp by to get the number of seconds and fractional seconds.

subsecond-field:

(v0.11.1+) The path to the property in a JSON-lines log message that contains the sub-second time value

subsecond-units:

(v0.11.1+) The units of the subsecond-field property value. The following values are supported:

milli:

for milliseconds

micro:

for microseconds

nano:

for nanoseconds

ordered-by-time:

(v0.8.3+) Indicates that the order of messages in the file is time-based. Files that are not naturally ordered by time will be sorted in order to display them in the correct order. Note that this sorting can incur a performance penalty when tailing logs.

level-field:

The name of the regex capture group that contains the log message level. Defaults to “level”.

body-field:

The name of the field that contains the main body of the message. Defaults to “body”.

opid-field:

The name of the field that contains the “operation ID” of the message. An “operation ID” establishes a thread of messages that might correspond to a particular operation/request/transaction. The user can press the ‘o’ or ‘Shift+O’ hotkeys to move forward/backward through the list of messages that have the same operation ID. Note: For JSON-encoded logs, the opid field can be a path (e.g. “foo/bar/opid”) if the field is nested in an object and it MUST be included in the “line-format” for the ‘o’ hotkeys to work.

module-field:

The name of the field that contains the module identifier that distinguishes messages from one log source from another. This field should be used if this message format can act as a container for other types of log messages. For example, an Apache access log can be sent to syslog instead of written to a file. In this case, lnav will parse the syslog message and then separately parse the body of the message to determine the “sub” format. This module identifier is used to help lnav quickly identify the format to use when parsing message bodies.

hide-extra:

A boolean for JSON logs that indicates whether fields not present in the line-format should be displayed on their own lines.

level:

A mapping of error levels to regular expressions. During scanning the contents of the capture group specified by level-field will be checked against each of these regexes. Once a match is found, the log message level will set to the corresponding level. The available levels, in order of severity, are: fatal, critical, error, warning, stats, info, debug, debug2-5, trace. For JSON logs with exact numeric levels, the number for the corresponding level can be supplied. If the JSON log format uses numeric ranges instead of exact numbers, you can supply a pattern and the number found in the log will be converted to a string for pattern-matching.

Note

The regular expression is not anchored to the start of the string by default, so an expression like 1 will match -1. If you want to exactly match 1, you would use ^1$ as the expression.

multiline:

If false, lnav will consider any log lines that do not match one of the message patterns to be in error when checking files with the ‘-C’ option. This flag will not affect normal viewing operation. Default: true.

value:

This object contains the definitions for the values captured by the regexes.

kind:

The type of data that was captured string, integer, float, json, quoted.

collate:

The name of the SQLite collation function for this value. The standard SQLite collation functions can be used as well as the ones defined by lnav, as described in Collators.

identifier:

A boolean that indicates whether or not this field represents an identifier and should be syntax colored.

foreign-key:

A boolean that indicates that this field is a key and should not be graphed. This should only need to be set for integer fields.

hidden:

A boolean for log fields that indicates whether they should be displayed. The behavior is slightly different for JSON logs and text logs. For a JSON log, this property determines whether an extra line will be added with the key/value pair. For text logs, this property controls whether the value should be displayed by default or replaced with an ellipsis.

rewriter:

A command to rewrite this field when pretty-printing log messages containing this value. The command must start with ‘:’, ‘;’, or ‘|’ to signify whether it is a regular command, SQL query, or a script to be executed. The other fields in the line are accessible in SQL by using the ‘:’ prefix. The text value of this field will then be replaced with the result of the command when pretty-printing. For example, the HTTP access log format will rewrite the status code field to include the textual version (e.g. 200 (OK)) using the following SQL query:

;SELECT :sc_status || ' (' || (
    SELECT message FROM http_status_codes
        WHERE status = :sc_status) || ') '
tags:

This object contains the tags that should automatically be added to log messages.

pattern:

The regular expression evaluated over a line in the log file as it is read in. If there is a match, the log message the line is a part of will have this tag added to it.

paths:

This array contains objects that define restrictions on the file paths that the tags will be applied to. The objects in this array can contain:

glob:

A glob pattern to check against the log files read by lnav.

sample:

A list of objects that contain sample log messages. All formats must include at least one sample and it must be matched by one of the included regexes. Each object must contain the following field:

line:

The sample message.

level:

The expected error level. An error will be raised if this level does not match the level parsed by lnav for this sample message.

highlights:

This object contains the definitions for patterns to be highlighted in a log message. Each entry should have a name and a definition with the following fields:

pattern:

The regular expression to match in the log message body.

color:

The foreground color to use when highlighting the part of the message that matched the pattern. If no color is specified, one will be picked automatically. Colors can be specified using hexadecimal notation by starting with a hash (e.g. #aabbcc) or using a color name as found at http://jonasjacek.github.io/colors/.

background-color:

The background color to use when highlighting the part of the message that matched the pattern. If no background color is specified, black will be used. The background color is only considered if a foreground color is specified.

underline:

If true, underline the part of the message that matched the pattern.

blink:

If true, blink the part of the message that matched the pattern.

Example format:

{
    "$schema": "https://lnav.org/schemas/format-v1.schema.json",
    "example_log" : {
        "title" : "Example Log Format",
        "description" : "Log format used in the documentation example.",
        "url" : "http://example.com/log-format.html",
        "regex" : {
            "basic" : {
                "pattern" : "^(?<timestamp>\\d{4}-\\d{2}-\\d{2}T\\d{2}:\\d{2}:\\d{2}\\.\\d{3}Z)>>(?<level>\\w+)>>(?<component>\\w+)>>(?<body>.*)$"
            }
        },
        "level-field" : "level",
        "level" : {
            "error" : "ERROR",
            "warning" : "WARNING"
        },
        "value" : {
            "component" : {
                "kind" : "string",
                "identifier" : true
            }
        },
        "sample" : [
            {
                "line" : "2011-04-01T15:14:34.203Z>>ERROR>>core>>Shit's on fire yo!"
            }
        ]
    }
}

Patching an Existing Format#

When loading log formats from files, lnav will overlay any new data over previously loaded data. This feature allows you to override existing value or append new ones to the format configurations. For example, you can separately add a new regex to the example log format given above by creating another file with the following contents:

{
    "$schema": "https://lnav.org/schemas/format-v1.schema.json",
    "example_log" : {
        "regex" : {
            "custom1" : {
                "pattern" : "^(?<timestamp>\\d{4}-\\d{2}-\\d{2}T\\d{2}:\\d{2}:\\d{2}\\.\\d{3}Z)<<(?<level>\\w+)--(?<component>\\w+)>>(?<body>.*)$"
            }
        },
        "sample" : [
            {
                "line" : "2011-04-01T15:14:34.203Z<<ERROR--core>>Shit's on fire yo!"
            }
        ]
    }
}

This example overrides the default syslog_log error detection regex to not match the errors= string.

{
  "syslog_log": {
      "level": {
          "error": "(?:(?:(?<![a-zA-Z]))(?:(?i)error(?:s)?(?!=))(?:(?![a-zA-Z]))|failed|failure)"
      }
  }
}

Scripts#

Format directories may also contain .sql and .lnav files to help automate log file analysis. The SQL files are executed on startup to create any helper tables or views and the ‘.lnav’ script files can be executed using the pipe hotkey |. For example, lnav includes a “partition-by-boot” script that partitions the log view based on boot messages from the Linux kernel. A script can have a mix of SQL and lnav commands, as well as include other scripts. The type of statement to execute is determined by the leading character on a line: a semi-colon begins a SQL statement; a colon starts an lnav command; and a pipe | denotes another script to be executed. Lines beginning with a hash are treated as comments. The following variables are defined in a script:

##

The number of arguments passed to the script.

__all__#

A string containing all the arguments joined by a single space.

0#

The path to the script being executed.

1-N#

The arguments passed to the script.

LNAV_HOME_DIR#

The path to the directory where the user’s lnav configuration is stored.

LNAV_WORK_DIR#

The path to the directory where lnav caches files, like archives that have been unpacked or piper captures.

Remember that you need to use the :eval command when referencing variables in most lnav commands. Scripts can provide help text to be displayed during interactive usage by adding the following tags in a comment header:

@synopsis:

The synopsis should contain the name of the script and any parameters to be passed. For example:

# @synopsis: hello-world <name1> [<name2> ... <nameN>]
@description:

A one-line description of what the script does. For example:

# @description: Say hello to the given names.

Tip

The :eval command can be used to do variable substitution for commands that do not natively support it. For example, to substitute the variable, pattern, in a :filter-out command:

:eval :filter-out ${pattern}

VSCode Extension#

The lnav VSCode Extension can be installed to add syntax highlighting to lnav scripts.

Installing Formats#

File formats are loaded from subdirectories in /etc/lnav/formats and ~/.lnav/formats/. You can manually create these subdirectories and copy the format files into there. Or, you can pass the ‘-i’ option to lnav to automatically install formats from the command-line. For example:

$ lnav -i myformat.json
info: installed: /home/example/.lnav/formats/installed/myformat_log.json

Format files installed using this method will be placed in the installed subdirectory and named based on the first format name found in the file.

You can also install formats from git repositories by passing the repository’s clone URL. A standard set of repositories is maintained at (https://github.com/tstack/lnav-config) and can be installed by passing ‘extra’ on the command line, like so:

lnav -i extra

These repositories can be updated by running lnav with the ‘-u’ flag.

Format files can also be made executable by adding a shebang (#!) line to the top of the file, like so:

#! /usr/bin/env lnav -i
{
    "myformat_log" : ...
}

Executing the format file should then install it automatically:

$ chmod ugo+rx myformat.json
$ ./myformat.json
info: installed: /home/example/.lnav/formats/installed/myformat_log.json

Format Order When Scanning a File#

When lnav loads a file, it tries each log format against the first 15,000 lines [1] of the file trying to find a match. When a match is found, that log format will be locked in and used for the rest of the lines in that file. Since there may be overlap between formats, lnav performs a test on startup to determine which formats match each others sample lines. Using this information it will create an ordering of the formats so that the more specific formats are tried before the more generic ones. For example, a format that matches certain syslog messages will match its own sample lines, but not the ones in the syslog samples. On the other hand, the syslog format will match its own samples and those in the more specific format. You can see the order of the format by enabling debugging and checking the lnav log file for the “Format order” message:

lnav -d /tmp/lnav.log

For JSON-lines log files, the log message must have the timestamp property specified in the format in order to match. If multiple formats match a message, the format that has the most matching line-format elements will win.

Automatic File Conversion#

File formats that are not naturally understood by lnav can be automatically detected and converted to a usable form using the converter property. For example, PCAP files can be detected and converted to a JSON-lines form using tshark. The conversion process works as follows:

  1. The first 1024 bytes of the file are read, if available.

  2. This header is converted into a hex string.

  3. For each log format that has defined a converter, every “header expression” is evaluated to see if there is a match. The header expressions are SQLite expressions where the following variables are defined:

    :header:

    A string containing the header as a hex string.

    :filepath:

    The path to the file.

  4. If a match is found, the converter script defined in the log format will be invoked and passed the format name and path to the file as arguments. The script should write the converted form of the input file on its standard output. Any errors should be written to the standard error.

  5. The log format will be associated with the original file will be used to interpret the converted file.