Cameron Katri's Manual Page Server

SENDING MESSAGES

The -s option is used send log messages to the syslogd(8) log message daemon, either locally or to a remote server if the -r host option in used.

There are two main forms of the command. If the -k option is used, then it must be followed by a list of keys and values. A structured message will be sent to the server with the keys and values given as arguments. If a key or a value has embedded white space, it must be enclosed in quotes.

Note that the text of the log message should be supplied as a value following the “Message” key.

If the -k option is not specified, then the rest of the command line is treated as the message text. The text may be preceded by -l level to set the log level (priority) of the message. Levels may be an integer value corresponding the the log levels specified in syslog(3) or asl(3), or they may be a string. String values are case insensitive, and should be one of:

Emergency

(level 0)

Alert

(level 1)

Critical

(level 2)

Error

(level 3)

Warning

(level 4)

Notice

(level 5)

Info

(level 6)

Debug

(level 7)

The string “Panic” is an alias for “Emergency”.

If the -l option is omitted, the log level defaults to 7 (Debug).

syslog only requires one or two leading characters for a level specification. A single character suffices in most cases. Use “P” or “Em” for Panic / Emergency, and “Er” or “X” for Error).

READING MESSAGES

The syslogd daemon filters and saves log messages to different output streams. One module saves messages to files specified in the syslog.conf(5) file. Those log files may be examined with any file printing or editing utility, e.g.

Another module saves messages in a data store (/var/log/asl).

If invoked with no arguments, syslog fetches all messages from the active data store. Messages are then printed to standard output, subject to formatting options and character encoding as described below. Some log messages are read-access controlled, so only messages that are readable by the user running syslog will be fetched and printed.

If invoked with the -C option, syslog fetches and prints console messages. The -C option is actually an alias for the expression:

See the EXPRESSIONS section below for more details.

Individual ASL data store files may be read by providing one or more file names as arguments to the -f option. This may be useful when searching archived files, files on alternate disk volumes, or files created as export files with the -x option.

The -d option may be followed by a list of directory paths. syslog will read or search all ASL data store files in those directories. Any files that are not readable will be skipped. Specifying -d with the name “archive” will open all readable files in the default ASL archive directory /var/log/asl.archive. Specifying -d with the name “store” will open all readable files in the ASL store directory /var/log/asl.

Legacy ASL database files that were written by syslogd on Mac OS X 10.5 (Leopard) may also be read using the -f option. However only one such legacy database may be read or searched at a time. Note that a legacy database may be read and copied into a new ASL data store format file using a combination of -f and -x options.

The -B option causes syslog to start processing messages beginning at the time of the last system startup. If used in conjunction with -w, all messages since the last system startup are displayed, or matched against an expression, before syslog waits for new messages.

The -w option causes syslog to wait for new messages. By default, syslog prints the last 10 messages, then waits for new messages to be added to the data store. A number following the -w option specifies the number of messages to print and overrides the default value of 10. For example:

Use the value “all” to view all messages in the data store before watching for new messages. The value “boot” will display messages since the last system startup before watching for new messages. Specifying “-w boot” is equivalent to using -w and -B together.

Using syslog with the -w option is similar to watching a log file using, e.g.

The -w option can only be used when reading the system's ASL data store or when reading a single data store file, and when printing messages to standard output.

If the -x file option is specified, messages are copied to the named file rather than being printed. The file will be created if it does not exist.

When called without the -x option, messages are printed to standard output. Messages are printed in a format similar to that used in the system.log file, except that the message priority level is printed between angle-brackets.

The output format may by changed by specifying the -F format option. Non-printable and control characters are encoded by default. Text encoding may be controlled using the -E option (see below). The value of format may be one of the following:

bsd

Format used by the syslogd daemon for system log files, e.g. /var/log/system.log.

std

Standard (default) format. Similar to “bsd”, but includes the message priority level.

raw

Prints the complete message structure. Each key/value pair is enclosed in square brackets. Embedded closing brackets and white space are escaped. Time stamps are printed as seconds since the epoch by default, but may also be printed in local time or UTC if the -T option is specified (see below).

xml

The list of messages is printed as an XML property list. Each message is represented as a dictionary in a array. Dictionary keys represent message keys. Dictionary values are strings.

Each of the format styles above may optionally be followed by a dot character and an integer value, for example:

This causes sub-second time values to be printed. In the example above, 4 decimal digits would be printed. The sub-second time values come from the value of the TimeNanoSec key in the ASL message. If the TimeNanoSec key is missing, a value of zero is used.

The value of the format argument may also be a custom print format string. A custom format should in most cases be enclosed in single quotes to prevent the shell from substituting special characters and breaking at white space.

Custom format strings may include variables of the form “$Name”, “$(Name)”, or “$((Name)(format))”. which will be expanded to the value associated with the named key. For example, the command:

produces output similar to the “std” format. The simple “$Name” form is sufficient in most cases. However, the second form: “$(Name)” must be used if the name is not delimited by white space. The third form allows different formats of the value to be printed. For example, a message priority level may appear as an integer value (e.g. “3”) or as a string (``Error''). The following print formats are known.

$((Level)(str))

Formats a Level value as a string, for example “Error”, “Alert”, “Warning”, and so on. Note that $(Level) or $Level formats the value as an integer 0 through 7.

$((Time)(sec))

Formats a Time value as the number of seconds since the Epoch.

$((Time)(raw))

Alias for $((Time)(sec)).

$((Time)(local))

Formats a Time value as a string of the form “Mmm dd hh:mm:ss”, where Mmm is the abbreviation for the month, dd is the date (1 - 31) and hh:mm:ss is the time. The local timezone is used.

$((Time)(lcl))

Alias for $((Time)(local)).

$((Time)(utc))

Formats a Time value as a string of the form “yyyy-mm-dd hh:mm:ssZ”, using Coordinated Universal Time, or the “Zulu” time zone.

$((Time)(zulu))

Alias for $((Time)(utc)).

$((Time)(X))

Where X may be any letter in the range A - Z or a - z. Formats the Time using the format “yyyy-mm-dd hh:mm:ssX”, using the specified nautical timezone. Z is the same as UTC/Zulu time. Timezones A - M (except J) decrease by one hour to the east of the Zulu time zone. Timezones N - Y increase by one hour to the west of Z. M and Y have the same clock time, but differ by one day. J is used to indicate the local timezone. When printing using $((Time)(J)), the output format is “yyyy-mm-dd hh:mm:ss”, without a trailing timezone letter.

$((Time)(JZ))

Specifies the local timezone. The timezone offset from UTC follows the date and time. The time is formatted as “yyyy-mm-dd hh:mm:ss[+|-]HH[:MM]”. Minutes in the timezone offset are only printed if they are non-zero.

$((Time)(ISO8601))

Specifies the local timezone and ISO 8601 extended format. The timezone offset from UTC follows the date and time. The time is formatted as “yyyy-mm-ddThh:mm:ss[+|-]HH[:MM]”. Minutes in the timezone offset are only printed if they are non-zero. Note that this differs from “JZ” format only in that a “T” character separates the date and time.

$((Time)(ISO8601B))

Specifies the local timezone and ISO 8601 basic format, in the form: “yyyymmddThhmmss[+|-]HH[:MM]”.

$((Time)(ISO8601Z))

Specifies UTC/Zulu time and ISO 8601 extended format, in the form: “yyyy-mm-ddThh:mm:ssZ”.

$((Time)(ISO8601BZ))

Specifies UTC/Zulu time and ISO 8601 basic format, in the form: “yyyymmddThhmmssZ”.

$((Time)([+|-]HH[:MM]))

Specifies an offset (+ or -) of the indicated number of hours (HH) and optionally minutes (MM) to UTC. The value is formatted as a string of the form “yyyy-mm-dd hh:mm:ss[+|-]HH[:MM]”. Minutes in the timezone offset are only printed if they are non-zero.

Each of the print formats listed above for Time values may optionally be followed by a dot character and an integer value. In that case, sub-second time values will be printed. For example, the following line prints messages with a UTC time format, and includes 6 digits of sub-second time:

If a custom format is not being used to specify the format for Time values, then Time values are generally converted to local time, except when the -F raw option is used, in which case times are printed as the number of seconds since the epoch. The -T format option may be used to control the format used for timestamps. The value of format may be one of the following:

sec or raw

Times are printed as the number of seconds since the epoch.

local or lcl

Times are converted to the local time zone, and printed with the format

mmm dd hh:mm:ss

where mmm is the month name abbreviated as three characters.

utc or zulu

Times are converted to UTC, and printed with the format

yyyy-mm-dd hh:mm:ssZ

A-Z

Times are converted to the indicated nautical time zone, printed in the same format as UTC. “J” is interpreted as the local timezone and printed in the same format, but without a trailing timezone letter.

JZ

is interpreted as the local timezone and printed with the format

yyyy-mm-dd hh:mm:ss[+|-]HH[:MM]

. The trailing “[+|-]HH[:MM]” string represents the local timezone offset from UTC in hours, or in hours and minutes if minutes are non-zero.

ISO8601

Times are printed with the format specified by ISO 8601:

yyyy-mm-ddThh:mm:ss[+|-]HH[:MM]

. This is the same as the “JZ” format, except a “T character separates the date and time components.”

[+|-]hh[:mm]

The specified offset is used to adjust time.

Each of the time formats above may optionally be followed by a dot character and an integer value. In that case, sub-second time values will be printed. For example:

The -u option is a short form for -T utc.

By default, control characters and non-printable characters are encoded in the output stream. In some cases this may make messages less natural in appearance. The encoding is designed to preserve all the information in the log message, and to prevent malicious users from spoofing or obscuring information in log messages.

Text in the “std”, “bsd”, and “raw” formats is encoded as it is by the vis utility with the -c option. Newlines and tabs are also encoded as "\n" and "\t" respectively. In “raw” format, space characters embedded in log message keys are encoded as "\s" and embedded brackets are escaped to print as "\[" and "\]".

XML format output requires that keys are valid UTF8 strings. Keys which are not valid UTF8 are ignored, and the associated value is not printed.

Values that contain legal UTF8 are printed as strings. Ampersand, less than, greater than, quotation mark, and apostrophe characters are encoded according to XML conventions. Embedded control characters are encoded as “&#xNN;” where NN is the character's hexadecimal value.

Values that do not contain legal UTF8 are encoded in base-64 and printed as data objects.

The -E format option may be used to explicitly control the text encoding. The value of format may be one of the following:

safe

This is the default encoding for syslog output. Encodes backspace characters as ^H. Carriage returns are mapped to newlines. A tab character is appended after newlines so that message text is indented.

vis

The C-style backslash encoding similar to that produced by the “vis -c” command, as described above.

none

No encoding is used.

The intent of the “safe” encoding is to prevent obvious message spoofing or damage. The appearance of messages printed will depend on terminal settings and UTF-8 string handling. It is possible that messages printed using the “safe” or “none” options may be garbled or subject to manipulation through the use of control characters and control sequences embedded in user-supplied message text. The “vis” encoding should be used to view messages if there is any suspicion that message text may have been used to manipulate the printed representation.

If no further command line options are specified, syslog displays all messages, or copies all messages to a data store file. However, an expression may be specified using the -k and -o options.

EXPRESSIONS

Expressions specify matching criteria. They may be used to search for messages of interest.

A simple expression has the form:

The -k option may be followed by one, two, or three arguments. A single argument causes a match to occur if a message has the specified key, regardless of value. If two arguments are specified, a match occurs when a message has exactly the specified value for a given key. For example, to find all messages sent by the portmap process:

Note that the -C option is treated as an alias for the expression:

This provides a quick way to search for console messages.

If three arguments are given, they are of the form -k key operation value. syslog supports the following matching operators:

eq

equal

ne

not equal

gt

greater than

ge

greater than or equal to

lt

less than

le

less than or equal to

Additionally, the operator may be preceded by one or more of the following modifiers:

C

case-fold

R

regular expression (see regex(3))

S

substring

A

prefix

Z

suffix

N

numeric comparison

More complex search expressions may be built by combining two or more simple expressions. A complex expression that has more than one “-k key [[op] val]” term matches a message if all of the key-value operations match. Logically, the result is an AND of all of key-value operations. For example:

finds all messages sent by portmap in the last 2 hours (-2h means "two hours ago").

The -o option may be used to build even more complex searches by providing an OR operation. If two or more sub-expressions are given, separated by -o options, then a match occurs is a message matches any of the sub-expressions. For example, to find all messages which have either a “Sender” value of “portmap” or that have a numeric priority level of 4 or less:

Log priority levels are internally handled as an integer value between 0 and 7. Level values in expressions may either be given as integers, or as string equivalents. See the table string values in the SENDING MESSAGES section for details. The example query above could also be specified with the command:

A special convention exists for matching time stamps. An unsigned integer value is regarded as the given number of seconds since 0 hours, 0 minutes, 0 seconds, January 1, 1970, Coordinated Universal Time. An negative integer value is regarded as the given number of seconds before the current time. For example, to find all messages of Error priority level (3) or less which were logged in the last 30 seconds:

a relative time value may be optionally followed by one of the characters “s”, “m”, “h”, “d”, or “w” to specify seconds, minutes, hours, days, or weeks respectively. Upper case may be used equivalently. A week is taken to be 7 complete days (i.e. 604800 seconds).

FILTERING CONTROLS

Clients of the Apple System Log facility using either the asl(3) or syslog(3) interfaces may specify a log filter mask. The mask specifies which messages should be sent to the syslogd daemon by specifying a yes/no setting for each priority level. Many clients set a filter mask to avoid sending relatively unimportant messages. Debug or Info priority level messages are generally only useful for debugging operations. By setting a filter mask, a process can improve performance by avoiding spending time sending messages that are in most cases unnecessary.

The -c option may be used to control filtering. In addition to the internal filter mask value that processes may set as described above, the system maintains a global “master” filter mask. This filter is normally “off”, meaning that it has no effect. If a value is set for the master filter mask, it overrides the local filter mask for all processes. Root user access is required to set the master filter mask value.

The current setting of the master filter mask may be inspected using:

The value of the master filter mask my be set by providing a second argument following -c 0. The value may a set of characters from the set “pacewnid”. These correspond to the priority levels Emergency (Panic), Alert, Critical, Error, Warning, Notice, Info, and Debug. The character “x” may be used for Error, as it is used for sending messages. The master filter mask may be deactivated with:

Since it is common to use the filter mask as a “cutoff” mechanism, for example to cut off messages with Debug and Info priority, a single character from the list above may be specified, preceded by a minus sign. In this case, syslog uses a filter mask starting at level 0 (Emergency) “up to” the given level. For example, to set the master filter mask to cause all processes to log messages from Emergency up to Debug:

While the master filter mask may be set to control the messages produced by all processes, another filter mask may be specified for an individual process. If a per-process filter mask is set, it overrides both the local filter mask and the master filter mask. The current setting for a per-process filter mask may be inspected using -c process, where process is either a PID or the name of a process. If a name is used, it must uniquely identify a process. To set a per-process filter mask, an second argument may be supplied following -c process as described above for the master filter mask. Root access is required to set the per-process filter mask for system (UID 0) processes.

The syslogd server follows filtering rules specified in the /etc/asl.conf file. When the remote-control mechanism is used to change the filter of a process, syslogd will save any messages received from that process until the remote-control filter is turned off.

SERVER CONFIGURATION

When syslogd starts up, and when it receives a HUP signal, it re-reads its configuration settings from /etc/asl.conf. It is sometimes useful to change configuration parameters temporarily, without needing to make changes to the configuration file. Any of the configuration options that may be set in the file (following an ``='' character) may also be sent to syslogd using the -config flag (without an ``='' character). For example, to temporarily disable the kernel message-per-second limit:

Note that only the superuser (root) may change configuration parameters.

In addition to the parameter setting options that are described in the asl.conf(5) manual page, an additional option:

will cause syslogd to reset its configuration.

ASL OUTPUT MODULES

ASL Output Modules are named configuration bundles used by the ASL server syslogd, and by the ASL filesystem manager aslmanager. The /etc/asl.conf file represents the system's primary output module, and is given the name “com.apple.asl”. Other modules are read from files in the /etc/asl directory. File names serve as module names. ASL Output Modules are described in detail in asl.conf(5).

When invoked with -module, syslog prints a summary of all loaded ASL Output Modules. The summary includes the output files and ASL store directories used by each module, a list of the module's configuration rules, and the module's current enabled or disabled status. -module name prints a summary for the module with the given name.

ASL Output Modules may be enabled or disabled using the command:

syslog -module name enable [0]

Note that only the superuser (root) may enable or disable a module.

The name '*' (including the single-quote characters) may be used to change the status of all ASL Output Modules, excluding the primary com.apple.asl module. com.apple.asl may be enabled or disabled, but only specifically by name.

If a module includes rotated files, the command:

syslog -module name checkpoint [file]

Will force the module to checkpoint all of its rotated files, or just the single optionally named file. The name '*' (including the single-quote characters) may be used to force checkpointing of all rotated files for all ASL Output Modules, including the primary com.apple.asl module.

Note that only the superuser (root) may force files to be checkpointed.

The checkpoint action sends a command to syslogd and waits for a reply to be returned. This means that any files currently in use will be checkpointed when the syslog command completes.