lnav/docs/source/data.rst
2015-07-22 07:09:51 -07:00

190 lines
9.4 KiB
ReStructuredText

.. _data-ext:
Extracting Data
===============
**Note**: This feature is still in **BETA**, you should expect bugs and
incompatible changes in the future.
Log messages contain a good deal of useful data, but it's not always easy to get
at. The log parser built into **lnav** is able to extract data as described by
:ref:`log-formats` as well as discovering data in plain text messages. This data
can then be queried and processed using the SQLite front-end that is also
incorporated into **lnav**. As an example, the following Syslog message from
:cmd:`sudo` can be processed to extract several key/value pairs::
Jul 31 11:42:26 Example-MacBook-Pro.local sudo[87024]: testuser : TTY=ttys004 ; PWD=/Users/testuser/github/lbuild ; USER=root ; COMMAND=/usr/bin/make install
The data that can be extracted by the parser is viewable directly in **lnav**
by pressing the 'p' key. The results will be shown in an overlay like the
following::
Current Time: 2013-07-31T11:42:26.000 Original Time: 2013-07-31T11:42:26.000 Offset: +0.000
Known message fields:
├ log_hostname = Example-MacBook-Pro.local
├ log_procname = sudo
├ log_pid = 87024
Discovered message fields:
├ col_0 = testuser
├ TTY = ttys004
├ PWD = /Users/testuser/github/lbuild
├ USER = root
└ COMMAND = /usr/bin/make install
Notice that the parser has detected pairs of the form '<key>=<value>'. The data
parser will also look for pairs separated by a colon. If there are no clearly
demarcated pairs, then the parser will extract anything that looks like data
values and assign them keys of the form 'col_N'. For example, two data values,
an IPv4 address and a symbol, will be extracted from the following log
messsage::
Apr 29 08:13:43 sample-centos5 avahi-daemon[2467]: Registering new address record for 10.1.10.62 on eth0.
Since there are no keys for the values in the message, the parser will assign
'col_0' for the IP address and 'col_1' for the symbol, as seen here::
Current Time: 2013-04-29T08:13:43.000 Original Time: 2013-04-29T08:13:43.000 Offset: +0.000
Known message fields:
├ log_hostname = sample-centos5
├ log_procname = avahi-daemon
├ log_pid = 2467
Discovered message fields:
├ col_0 = 10.1.10.62
└ col_1 = eth0
Now that you have an idea of how the parser works, you can begin to perform
queries on the data that is being extracted. The SQLite database engine is
embedded into **lnav** and its `Virtual Table
<http://www.sqlite.org/vtab.html>`_ mechanism is used to provide a means to
process this log data. Each log format has its own table that can be used to
access all of the loaded messages that are in that format. For accessing log
message content that is more free-form, like the examples given here, the
**logline** table can be used. The **logline** table is recreated for each
query and is based on the format and pairs discovered in the log message at
the top of the display.
Queries can be performed by pressing the semi-colon (;) key in **lnav**. After
pressing the key, the overlay showing any known or discovered fields will be
displayed to give you an idea of what data is available. The query can be any
`SQL query <http://sqlite.org/lang.html>`_ supported by SQLite. To make
analysis easier, **lnav** includes many extra functions for processing strings,
paths, and IP addresses. See :ref:`sql-ext` for more information.
As an example, the simplest query to perform initially would be a "select all",
like so::
select * from logline
When this query is run against the second example log message given above, the
following results are received::
log_line log_part log_time log_idle_msecs log_level log_hostname log_procname log_pid col_0 col_1
292 p.0 2013-04-11T16:42:51.000 0 info localhost avahi-daemon 2480 fe80::a00:27ff:fe98:7f6e eth0
293 p.0 2013-04-11T16:42:51.000 0 info localhost avahi-daemon 2480 10.0.2.15 eth0
330 p.0 2013-04-11T16:47:02.000 0 info localhost avahi-daemon 2480 fe80::a00:27ff:fe98:7f6e eth0
336 p.0 2013-04-11T16:47:02.000 0 info localhost avahi-daemon 2480 10.1.10.75 eth0
343 p.0 2013-04-11T16:47:02.000 0 info localhost avahi-daemon 2480 10.1.10.75 eth0
370 p.0 2013-04-11T16:59:39.000 0 info localhost avahi-daemon 2480 10.1.10.75 eth0
377 p.0 2013-04-11T16:59:39.000 0 info localhost avahi-daemon 2480 10.1.10.75 eth0
382 p.0 2013-04-11T16:59:41.000 0 info localhost avahi-daemon 2480 fe80::a00:27ff:fe98:7f6e eth0
401 p.0 2013-04-11T17:20:45.000 0 info localhost avahi-daemon 4247 fe80::a00:27ff:fe98:7f6e eth0
402 p.0 2013-04-11T17:20:45.000 0 info localhost avahi-daemon 4247 10.1.10.75 eth0
735 p.0 2013-04-11T17:41:46.000 0 info sample-centos5 avahi-daemon 2465 fe80::a00:27ff:fe98:7f6e eth0
736 p.0 2013-04-11T17:41:46.000 0 info sample-centos5 avahi-daemon 2465 10.1.10.75 eth0
781 p.0 2013-04-12T03:32:30.000 0 info sample-centos5 avahi-daemon 2465 10.1.10.64 eth0
788 p.0 2013-04-12T03:32:30.000 0 info sample-centos5 avahi-daemon 2465 10.1.10.64 eth0
1166 p.0 2013-04-25T10:56:00.000 0 info sample-centos5 avahi-daemon 2467 fe80::a00:27ff:fe98:7f6e eth0
1167 p.0 2013-04-25T10:56:00.000 0 info sample-centos5 avahi-daemon 2467 10.1.10.111 eth0
1246 p.0 2013-04-26T06:06:25.000 0 info sample-centos5 avahi-daemon 2467 10.1.10.49 eth0
1253 p.0 2013-04-26T06:06:25.000 0 info sample-centos5 avahi-daemon 2467 10.1.10.49 eth0
1454 p.0 2013-04-28T06:53:55.000 0 info sample-centos5 avahi-daemon 2467 10.1.10.103 eth0
1461 p.0 2013-04-28T06:53:55.000 0 info sample-centos5 avahi-daemon 2467 10.1.10.103 eth0
1497 p.0 2013-04-29T08:13:43.000 0 info sample-centos5 avahi-daemon 2467 10.1.10.62 eth0
1504 p.0 2013-04-29T08:13:43.000 0 info sample-centos5 avahi-daemon 2467 10.1.10.62 eth0
Note that **lnav** is not returning results for all messages that are in this
syslog file. Rather, it searches for messages that match the format for the
given line and returns only those messages in results. In this case, that
format is "Registering new address record for <IP> on <symbol>", which
corresponds to the parts of the message that were not recognized as data.
More sophisticated queries can be done, of course. For example, to find out the
frequency of IP addresses mentioned in these messages, you can run::
SELECT col_0,count(*) FROM logline GROUP BY col_0
The results for this query are::
col_0 count(*)
10.0.2.15 1
10.1.10.49 2
10.1.10.62 2
10.1.10.64 2
10.1.10.75 6
10.1.10.103 2
10.1.10.111 1
fe80::a00:27ff:fe98:7f6e 6
Since this type of query is fairly common, **lnav** includes a "summarize"
command that will compute the frequencies of identifiers as well as min, max,
average, median, and standard deviation for number columns. In this case, you
can run the following to compute the frequencies and return an ordered set of
results.
:summarize col_0
Recognized Data Types
---------------------
When searching for data to extract from log messages, **lnav** looks for the
following set of patterns:
Strings
Single and double-quoted strings. Example: "The quick brown fox."
URLs
URLs that contain the '://' separator. Example: http://example.com
Paths
File system paths. Examples: /path/to/file, ./relative/path
MAC Address
Ethernet MAC addresses. Example: c4:2c:03:0e:e4:4a
Hex Dumps
A colon-separated string of hex numbers. Example: e8:06:88:ff
Date/Time
Date and time stamps of the form "YYYY-mm-DD" and "HH:MM:SS".
IP Addresses
IPv4 and IPv6 addresses. Examples: 127.0.0.1, fe80::c62c:3ff:fe0e:e44a%en0
UUID
The common formatting for 128-bit UUIDs. Example:
0E305E39-F1E9-4DE4-B10B-5829E5DF54D0
Version Numbers
Dot-separated version numbers. Example: 3.7.17
Numbers
Numbers in base ten, hex, and octal formats. Examples: 1234, 0xbeef, 0777
E-Mail Address
Strings that look close to an e-mail address. Example: gary@example.com
Constants
Common constants in languages, like: true, false, null, None.
Symbols
Words that follow the common conventions for symbols in programming
languages. For example, containing all capital letters, or separated
by colons. Example: SOME_CONSTANT_VALUE, namespace::value