Logging Reference
Logging in Riak KV is handled by a Basho-produced logging framework for Erlang called lager.
lager provides a number of configuration options that you can use to fine-tune your Riak cluster’s logging output. A compact listing of parameters can be found in our configuration files documentation. A more thorough explanation of these options can be found in this document.
Log Directory
Riak’s log files are stored in a /log
directory on each node. The
location of that directory differs from platform to platform. The table
below shows you where log files are stored on all supported operating
systems.
OS | Directory |
---|---|
Ubuntu, Debian, CentOS, RHEL | /var/log/riak |
Solaris, OpenSolaris | /opt/riak/log |
Source install and Mac OS X | ./log (where the . represents the root installation directory) |
Log Files
Below is a list of files that can be found in each node’s /log
directory:
File | Significance |
---|---|
console.log |
Console log output |
crash.log |
Crash logs |
erlang.log |
Logs emitted by the Erlang VM on which Riak runs. |
error.log |
Common errors emitted by Riak. |
run_erl.log |
The log file for an Erlang process called run_erl . This file can typically be ignored. |
Log Syntax
Riak logs tend to be structured like this:
<date> <time> [<level>] <PID> <prefix>: <message>
The date
segment is structured YYYY-MM-DD
, time
is structured
hh:mm:ss.sss
, level
depends on which log levels are available in the
file you are looking at (consult the sections below), the PID
is the
Erlang process identifier for the process in which the event occurred,
and the message prefix
will often identify the Riak subsystem
involved, e.g. riak_ensemble_peer
or alarm_handler
(amongst many
other possibilities).
As of Riak KV 3.0.3 a few of the log messages may contain newline characters, preventing reliable identification of the end of each log when attempting log files ingestion by external tools.
A known workaround is ingesting not the logs enabled by the
log.console
configurable parameter but rather the logs as enabled by
the log.syslog
configurable parameter and processed by syslog,
e.g. exploiting the
no-multi-line
option (e.g. see this StackExchange topic
answer)
- or equivalent - of syslog implementations.
The exception to this syntax is in crash logs (stored in crash.log
files). For crash logs, the syntax tends to be along the following
lines:
<date> <time> =<report title>====
<message>
Here is an example crash report:
2014-10-17 15:56:38 =ERROR REPORT====
Error in process <0.4330.323> on node 'dev1@127.0.0.1' with exit value: ...
Log Files
In each node’s /log
directory, you will see at least one of each of
the following:
File | Contents |
---|---|
console.log |
General messages from all Riak subsystems |
crash.log |
Catastrophic events, such as node failures, running out of disk space, etc. |
erlang.log |
Events from the Erlang VM on which Riak runs |
run_erl.log |
The command-line arguments used when starting Riak |
Log File Rotation
Riak maintains multiple separate files for console.log
, crash.log
,
erlang.log
, and error.log
, which are rotated as each file reaches
its maximum capacity of 100 KB. In each node’s /log
directory, you may
see, for example, files name console.log
, console.log.0
,
console.log.1
, and so on. Riak’s log rotation is somewhat non
traditional, as it does not always log to *.1
(e.g. erlang.log.1
)
but rather to the oldest log file.
After, say, erlang.log.1
is filled up, the logging system will begin
writing to erlang.log.2
, then erlang.log.3
, and so on. When
erlang.log.5
is filled up, it will loop back to erlang.log.1
.
SASL
SASL (System Architecture
Support Libraries) is Erlang’s built-in error logger. You can enable it
and disable it using the sasl
parameter (which can be set to on
or
off
). It is disabled by default. The following would enable it:
sasl = on
Error Messages
By default, Riak stores error messages in ./log/error.log
by default.
You can change this using the log.error.file
parameter. Here is an
example, which uses the default:
log.error.file = ./log/error.log
By default, error messages are redirected into lager, i.e. the
log.error.redirect
parameter is set to on
. The following would
disable the redirect:
log.error.redirect = off
You can also throttle the number of error messages that are handled per second. The default is 100.
log.error.messages_per_second = 100
Crash Logs
Riak crash logs are stored in ./log/crash.log
by default. You can
change this using the log.crash.file
parameter. This example uses the
default:
log.crash.file = ./log/crash.log
While crash logs are kept by default, i.e. the log.crash
parameter is
set to on
, you can disable crash logs like this:
log.crash = off
Crash Log Rotation
Like other Riak logs, crash logs are rotated. You can set the crash logs to be rotated either when a certain size threshold is reached and/or at designated times.
You can set the rotation time using the log.crash.rotation
parameter.
The default is $D0
, which rotates the logs every day at midnight. You
can also set the rotation to occur weekly, on specific days of the
month, etc. Complete documentation of the syntax can be found
here.
Below are some examples:
$D0
- Every night at midnight$D23
- Every day at 23:00 (11 pm)$W0D20
- Every week on Sunday at 20:00 (8 pm)$M1D0
- On the first day of every month at midnight$M5D6
- On the fifth day of the month at 6:00 (6 am)
To set the maximum size of the crash log before it is rotated, use the
log.crash.size
parameter. You can specify the size in KB, MB, etc. The
default is 10MB
.
Other Crash Log Settings
The maximum size of individual crash log messages can be set using the
log.crash.maximum_message_size
, using any size denomination you wish,
e.g. KB
or MB
The default is 64 KB. The following would set that
maximum message size to 1 MB:
log.crash.maximum_message_size = 1MB
Syslog
Riak log output does not go to syslog by default, i.e. the log.syslog
setting is set to off
by default. To enable syslog output:
log.syslog = on
If syslog output is enabled, you can choose a prefix to be appended to
each syslog message. The prefix is riak
by default.
log.syslog.ident = riak
Syslog Level and Facility Level
If syslog is enabled, i.e. if log.syslog
is set to on
, you can
select the log level of syslog output from amongst the available levels,
which are listed in the table below. The default is info
.
alert
critical
debug
emergency
error
info
none
notice
warning
In addition to a log level, you must also select a facility
level for syslog
messages amongst the available levels, which are listed in the table
below. The default is daemon
.
auth
authpriv
clock
cron
daemon
ftp
kern
lpr
mail
news
syslog
user
uucp
In addition to these options, you may also choose one of local0
through local7
.
Console Logs
Riak console logs can be emitted to one of three places: to a log file
(you can choose the name and location of that file), to standard output,
or to neither. This is determined by the value that you give to the
log.console
parameter, which gives you one of four options:
file
- Console logs will be emitted to a file. This is Riak’s default behavior. The location of that file is determined by thelog.console.file
parameter. The default location is./log/console.log
on an installation from source, but will differ on platform-specific installation, e.g./var/log/riak
on Ubuntu, Debian, CentOS, and RHEL or/opt/riak/log
on Solaris-based platforms.console
- Console logs will be emitted to standard output, which can be viewed by running theriak attach-direct
commandboth
- Console logs will be emitted both to a file and to standard outputoff
- Console log messages will be disabled
In addition to the the placement of console logs, you can also choose
the severity of those messages using the log.console.level
parameter.
The following four options are available:
info
(the default)debug
warning
error
Enabling and Disabling Debug Logging
Checkout Cluster Operations: Enabling and Disabling Debug Logging