Archives/lnav
2
0
Fork 0
mirror of https://github.com/tstack/lnav synced 2024-11-17 15:29:40 +00:00
Code Issues Packages Projects Releases Wiki Activity
7eeb45f5e8
lnav/src/help.txt

508 lines
22 KiB
Plaintext
Raw Blame History

lnav - A fancy log file viewer
DESCRIPTION
-----------
The log file navigator, lnav, is an enhanced log file viewer that
takes advantage of any semantic information that can be gleaned from
the files being viewed, such as timestamps and log levels. Using this
extra semantic information, lnav can do things like interleaving
messages from different files, generate histograms of messages over
time, and providing hotkeys for navigating through the file. It is
hoped that these features will allow the user to quickly and
efficiently zero in on problems.
OPTIONS
-------
Lnav takes a list of files to view and/or you can use the flag
arguments to load well-known log files, such as the syslog or apache
log files. The flag arguments are:
-s Load the most recent syslog messages file. (Default)
-a Load all of the most recent log file types.
-r Load older rotated log files as well.
When using the flag arguments, lnav will look for the files relative
to the current directory and its parent directories. In other words,
if you are working within a directory that has the well-known log
files, those will be preferred over any others.
Any files given on the command-line are scanned to determine their log
file format and to create an index for each line in the file. You do
not have to manually specify the log file format. The currently
supported formats are: syslog, apache, strace, tcsh history, and
generic log files with timestamps.
Lnav will also display data piped in on the standard input. The
following options are available when doing so:
-t Prepend timestamps to the lines of data being read in
on the standard input.
-w file Write the contents of the standard input to this file.
DISPLAY
-------
The main part of the display shows the log lines from the files interleaved
based on time-of-day. New lines are automatically loaded as they are appended
to the files and, if you are viewing the bottom of the files, lnav will scroll
down to display the new lines, much like 'tail -f'.
On color displays, the lines will be highlighted as follows:
* Errors will be colored in red;
* warnings will be yellow;
* boundaries between days will be underlined; and
* various color highlights will be applied to: IP addresses, SQL keywords,
XML tags, file and line numbers in Java backtraces, and quoted strings.
To give you an idea of where you are in the file spatially, the right
side of the display has a proportionally sized 'scrollbar' that
indicates your current position in the file.
Above and below the main body are status lines that display:
* the current time;
* the name of the file the top line was pulled from;
* the log format for the top line;
* the current view;
* the line number for the top line in the display;
* the total number of warnings and errors;
* the number of search hits, which updates as more are found; and
* the number of lines not displayed because of filtering.
Finally, the last line on the display is where you can enter search
patterns and execute internal commands, such as converting a
unix-timestamp into a human-readable date. The command-line is by
the readline library, so the usual set of keyboard shortcuts can
be used.
The body of the display is also used to display other content, such
as: the help file, histograms of the log messages over time, and
SQL results. The views are organized into a stack so that any time
you activate a new view with a key press or command, the new view
is pushed onto the stack. Pressing the same key again will pop the
view off of the stack and return you to the previous view. Note
that you can always use 'q' to pop the top view off of the stack.
KEY BINDINGS
------------
To help navigate through the file there are many hotkeys that should
make it easy to zero-in on a specific section of the file or scan
through the file.
? View/leave this help message.
q Leave the current view or quit the program when in
the log file view.
home Move to the top of the file.
end Move to the end of the file. If the view is already
at the end, it will move to the last line.
space/pgdn Move down a page.
b/bs/pgup Move up a page.
j/cr/down-arrow Move down a line.
k/up-arrow Move up a line.
h/left-arrow Move to the left.
l/right-arrow Move to the right.
e/E Move to the next/previous error.
w/W Move to the next/previous warning.
n/N Move to the next/previous search hit.
f/F Move to the next/previous file. In the log view, this
moves to the next line from a different file. In the
text view, this rotates the view to the next file.
>/< Move horizontally to the next/previous search hit.
s Toggle "scrubbing" of log lines in the log view. Log
formats that support scrubbing will try to eliminate
redundant information on a single line. For example, the
year part of a timestamp or the log level text.
t Switch to/from the text file view. The text file view is
for any files that are not recognized as log files.
o/O Move forward/backward 60 minutes from the current
position in the log file.
d/D Move forward/backward 24 hours from the current
position in the log file.
1-6/Shift 1-6 Move to the next/previous n'th ten minute of the
hour. For example, '4' would move to the first
log line in the fortieth minute of the current
hour in the log. And, '6' would move to the next
hour boundary.
0/Shift 0 Move to the next/previous day boundary.
m Mark/unmark the line at the top of the display.
The line will be highlighted with reverse video to
indicate that it is a user bookmark. You can use
the 'u' hotkey to iterate through marks you have
added.
M Mark/unmark all the lines between the top of the
display and the last line marked/unmarked.
J Mark/unmark the next line after the previously
marked line.
K Like 'J' except it toggles the mark on the
previous line.
c Copy the marked text to the X11 selection buffer or OS X
clipboard.
C Clear all marked lines.
u/U Move forward/backward through any user bookmarks
you have added using the 'm' key.
T Toggle the display of the "time offset" column that shows
the time elapsed since the beginning of the logs or the
offset from the previous bookmark.
i View/leave a histogram of the log messages over
time. The histogram counts the number of
displayed log lines for each bucket of time. The
bars are layed out horizontally with colored
segments representing the different log levels.
You can use the 'z' hotkey to change the size of
the time buckets (e.g. ten minutes, one hour, one
day).
I Switch between the log and histogram views while
keeping the time displayed at the top of each view
in sync. For example, if the top line in the log
view is "11:40", hitting 'I' will switch to the
histogram view and scrolled to display "11:00" at
the top (if the zoom level is hours).
z/Shift Z Zoom in or out one step in the histogram view.
/<regexp> Start a search for the given regular expression.
The search is live, so when there is a pause in
typing, the currently running search will be
canceled and a new one started. History is
maintained for your searches so you can rerun them
easily. If there is an error encountered while
trying to interpret the expression, the error will
be displayed in red on the status line. While the
search is active, the 'hits' field in the status
line will be green, when finished it will turn
back to black.
Note: The regular expression format used by is PCRE
(Perl-Compatible Regular Expressions). For example,
if you wanted to search for ethernet device names,
regardless of their ID number, you can type:
eth\d+
You can find more information about Perl regular
expressions at:
http://perldoc.perl.org/perlre.html
:<command> Execute an internal command. The commands are
listed below. History is also supported in this
context as well as tab-completion for commands and
some arguments. The result of the command
replaces the command you typed.
;<sql> Execute an SQL query. Most supported log file
formats provide a sqlite virtual table backend
that can be used in queries. See the SQL section
below for more information.
CTRL+] Abort command-line entry started with '/', ':', or ';'.
y/Y Move forward/backward through the log view based on the
"log_line" column in the SQL result view.
v Switch to/from the SQL result view.
V Switch between the log and SQL result views while
keeping the top line number in the log view in
sync with the log_line column in the SQL view.
For example, doing a query that selects for
"log_idle_msecs" and "log_line", you can move the
top of the SQL view to a line and hit 'V' to switch
to the log view and move to the line number that was
selected in the "log_line" column.
TAB/Shift+TAB In the SQL result view, cycle through the columns that
are graphed. Initially, all number values are displayed
in a stacked graph. Pressing TAB will change the display
to only graph the first column. Repeatedly pressing TAB
will cycle through the columns until they are all graphed
again.
p Enable or disable the display of the fields that the
log message parser knows about or has discovered.
This overlay is temporarily enabled when the semicolon
key (;) is pressed so that it is easier to write queries.
X Close the current text file or log file.
CTRL-R Reset the session state. This will save the current
session state (filters, highlights) and then reset the
state to the factory default.
r/R Restore the next/previous session. The current session is
saved and then the new state is restored.
F2 Toggle mouse support.
MOUSE SUPPORT (experimental)
----------------------------
If you are using Xterm, or a compatible terminal, you can use the mouse to
mark lines of text and move the view by grabbing the scrollbar.
NOTE: You need to manually enable this feature by setting the LNAV_EXP
environment variable to "mouse".
COMMANDS
--------
help Switch to this help text view.
adjust-log-time <date>
Change the time of the top log line to the given time.
All other log lines in the same file will also be
adjusted using the same offset. After the adjustment,
the displayed timestamp will be rewritten to the new
time and highlighted with a magenta color.
This command is useful for lining up log files that
have timestamps from different machines.
unix-time <secs-or-date>
Convert a unix-timestamp in seconds to a
human-readable form or vice-versa.
BEWARE OF TIMEZONE DIFFERENCES.
current-time Print the current time in human-readable form and
as a unix-timestamp.
goto <line#|N%> Go to the given line number or N percent into the
file.
highlight <regex> Highlight strings that match the given regular
expression.
filter-in <regex> Only display lines that match the given regular
expression. This command can be used multiple
times to add more lines to the display. The number
of lines that are filtered out will be shown in the
bottom status bar as 'Not Shown'.
filter-out <regex>
Do not display lines that match the given regular
expression. This command can be used multiple
times to remove more lines from the display. If a
'filter-in' expression is also active, it takes
priority and the filter-out will remove lines that
were matched by the 'filter-in'. The number
of lines that are filtered out will be shown in the
bottom status bar as 'Not Shown'.
disable-filter <regex>
Disable an active 'filter-in' or 'filter-out'
expression.
enable-filter <regex>
Enable a inactive 'filter-in' or 'filter-out'
expression.
disable-word-wrap Disable word wrapping in the log and text file views.
enable-word-wrap Enable word wrapping in the log and text file views.
open <filename>[:<line>]
Open the given file within lnav and, if it is a
text file, switch to the text view and jump to
the given line number.
close Close the current text file or log file. You can also
close the current file by pressing 'X'.
graph <regex> Graph the value of numbers in the file(s) over
time. The given regular expression should capture
the number to be displayed. For example:
my stats: (\d+\.\d+)
Will graph all the "stats" values found in the
file. XXX This is still mostly a toy...
append-to <file> Append any marked lines to the given file.
write-to <file> Write any marked lines to the given file.
session <cmd> Add the given command to the session file
(~/.lnav/session). Any commands listed in the session file
are executed on startup. Only the highlight, word-wrap, and
filter-related commands can be added to the session file.
create-logline-table <table-name>
Create an SQL table using the top line of the log view
as a template. See the "SQL QUERIES" and "DYNAMIC LOG
LINE TABLE" sections below for more information.
delete-logline-table <table-name>
Delete an SQL table created by the 'create-logline-table'
command.
SQL QUERIES (experimental)
-----------
Lnav has support for performing SQL queries on log files using the
Sqlite3 "virtual" table feature. For all supported log file types,
lnav will create tables that can be queried using the subset of SQL
that is supported by Sqlite3. For example, to get the top ten URLs
being accessed in any loaded Apache log files, you can execute:
;select cs_uri_stem, count(*) as total from access_log
group by cs_uri_stem order by total desc limit 10;
The query result view shows the results as text and graphs any number
values found in the result, much like the histogram view.
The builtin set of log tables are listed below. Note that only the
log messages that match a particular format can be queried by a
particular table. You can find the file format and table name for
the top log message by looking in the upper right hand corner of the
log file view.
The log table names are as follows:
access_log Apache common access log format
syslog_log Syslog format
glog_log Google glog format
strace_log Strace log format
generic_log 'Generic' log format. This table contains messages
from files that have a very simple format with a
leading timestamp followed by the message.
The columns available for the top log line in the view will
automatically be displayed after pressing the semicolon (;) key.
All log tables contain at least the following columns:
log_line The line number in the file, starting at zero.
log_time The time of the log entry.
log_idle_msecs The amount of time, in milliseconds, between the
current log message and the previous one.
log_level The log level (e.g. info, error, etc...).
log_path The full path to the file.
log_text The raw line of text. Note that this column is
not included in the result of a 'select *', but
it does exist.
The following tables include the basic columns as listed above and
include a few more columns since the log file format is more
structured.
syslog_log
log_hostname The hostname the message was received from.
log_procname The name of the process that sent the message.
log_pid The process ID of the process that sent the message.
access_log (The column names are the same as those in the
Microsoft LogParser tool.)
c_ip The client IP address.
cs_username The client user name.
cs_method The HTTP method.
cs_uri_stem The stem portion of the URI.
cs_uri_query The query portion of the URI.
cs_version The HTTP version string.
sc_status The status number returned to the client.
sc_bytes The number of bytes sent to the client.
cs_referrer The URL of the referring page.
cs_user_agent The user agent string.
strace_log (Currently, you need to run strace with the
"-tt -T" options so there are timestamps for
each function call.)
funcname The name of the syscall.
result The result code.
duration The amount of time spent in the syscall.
arg0 - arg9 The arguments passed to the syscall.
These tables are created dynamically and not stored in memory or on
disk. If you would like to persist some information from the tables,
you can attach another database and create tables in that database.
For example, if you wanted to save the results from the earlier
example of a top ten query into the "/tmp/topten.db" file, you can do:
;attach database "/tmp/topten.db" as topten;
;create table topten as select cs_uri_stem, count(*) as total
from access_log group by cs_uri_stem order by total desc
limit 10;
DYNAMIC LOG LINE TABLE (experimental)
----------------------
(NOTE: This feature is still very new and not completely reliable yet,
use with care.)
For log formats that lack message structure, lnav can parse the log
message and attempt to extract any data fields that it finds. This
feature is available through the "logline" log table. This table is
dynamically created and defined based on the message at the top of
the log view. For example, given the following log message from "sudo",
lnav will create the "logline" table with columns for "TTY", "PWD",
"USER", and "COMMAND":
May 24 06:48:38 Tim-Stacks-iMac.local sudo[76387]: stack : TTY=ttys003 ;
PWD=/Users/stack/github/lbuild ; USER=root ;
COMMAND=/bin/echo Hello, World!
Queries executed against this table will then only return results for
other log messages that have the same format. So, if you were to
execute the following query while viewing the above line, you might
get the following results:
;select USER,COMMAND from logline;
USER | COMMAND
---- | -------------------------
root | /bin/echo Hello, World!
mal | /bin/echo Goodbye, World!
The log parser works by examining each message for key/value pairs
separated by an equal sign (=) or a colon (:). For example, in the
previous example of a "sudo" message, the parser sees the "USER=root"
string as a pair where the key is "USER" and the value is "root".
If no pairs can be found, then anything that looks like a value is
extracted and assigned a numbered column. For example, the following
line is from "dhcpd":
Sep 16 22:35:57 drill dhcpd: DHCPDISCOVER from 00:16:ce:54:4e:f3 via hme3
In this case, the lnav parser recognizes that "DHCPDISCOVER", the MAC
address and the "hme3" device name are values and not normal words. So,
it builds a table with three columns for each of these values. The
regular words in the message, like "from" and "via", are then used to
find other messages with a similar format.
If you would like to execute queries against log messages of different
formats at the same time, you can use the 'create-logline-table' command
to permanently create a table using the top line of the log view as a
template.
Reference in New Issue View Git Blame Copy Permalink