logging.yaml¶

The logging.yaml file defines all custom log file formats, filters, and processing options.

重要

This configuration file replaces the XML based logs_xml.config, as well as the Lua based logging.config from past Traffic Server releases. If you are upgrading from a Traffic Server release which used either the XML or the Lua configuration file format, and you have created custom log formats, filters, and destinations, you will need to update those settings to this format.

Log Definitions¶

Custom logs are configured by the combination of three key elements: a format, an optional filter, and a log destination.

A format defines how log lines will appear (as well as whether the logs using the format will be event logs or summary logs).

A filter defines what events do, and what events don't, make it into the logs employing the filter.

A log defines where the record of events or summaries ends up.

Formats¶

Custom logging formats may be provided directly to a log definition, or they may be defined as a reusable variable in your logging.yaml for ease of reference, particularly when you may have more than one log using the same format. Which approach you use is entirely up to you, though it's strongly recommended to create an explicit format object if you intend to reuse the same format for multiple log files.

Custom formats are defined by choosing a name to identify the given logging format, and a format string, which defines the output format string for every event. An optional interval attribute can be specified to define the aggregation interval for summary logs.

# A one-line-per-event format that just prints event timestamps.
formats:
- name: myformat
  format: '%<cqtq>'

# An aggregation/summary format that prints the last event timestamp from
# the interval along with the total count of events in the same interval.
# (Doing so every 30 seconds.)
formats:
- name: mysummaryformat
  format: '%<LAST(cqtq)> %<COUNT(*)>'
  interval: 30

You may define as many and as varied a collection of format objects as you desire.

Format Specification¶

The format specification provided as the required format attribute of the objects listed in formats is a simple string, containing whatever mixture of logging field variables and literal characters meet your needs. Logging fields are discussed in great detail in the Log Fields section.

Flexible enough to not only emulate the logging formats of most other proxy and HTTP servers, but also to provide even finer detail than many of them, the logging fields are very easy to use. Within the format string, logging fields are indicated by enclosing their name within angle brackets (< and >), preceded by a percent symbol (%). For example, returning to the altogether too simple format shown earlier, the following format string:

'%<cqtq>'

Defines a format in which nothing but the value of the logging field cqtq is interpolated for each event's entry in the log. We could include some literal characters in the log output by updating the format specification as so:

'Event received at %<cqtq>'

Because the string "Event received at " (including the trailing space) is just a bunch of characters, not enclosed in %<...>, it is repeated verbatim in the logging output.

Multiple logging fields may of course be used:

'%<cqtq> %<chi> %<cqhm> %<pqu> %<cqpv>'

Each logging field is separately enclosed in its own percent-brace set.

There are a small number of logging fields which extend this simple format, primarily those dealing with request and response headers. Instead of defining a separate logging field name for every single possible HTTP header (an impossible task, given that arbitrary vendor/application headers may be present in both requests and responses), there are instead single logging fields for each of the major stages of an event lifecycle that permit access to named headers, such as:

'%<{User-Agent}cqh>'

Which emits to the log the value of the client request's User-Agent HTTP header. Other stages of the event lifecycle have similar logging fields: pqh (proxy requests), ssh (origin server responses), and psh (proxy responses).

You will find a complete listing of the available fields in Log Fields.

Aggregation Interval¶

Every format may be given an optional interval value, specified as the number of seconds over which events destined for a log using the format are aggregated and summarized. Logs which use formats containing an aggregation interval do not behave like regular logs, with a single line for every event. Instead, they emit a single line only every interval-seconds.

These types of logs are described in more detail in Summary Logs.

Formats have no interval by default, and will generate event-based logs unless given one.

Filters¶

Trafficserver supports different type of filters : accept, reject and wipe_field_value. They may be used, optionally, to accept, reject logging or mask query param values for matching events.

Filter objects are created by assigning them a name to be used later to refer to the filter, as well as an action (either accept, reject or wipe_field_value). Accept, reject or wipe_field_value filters require a condition against which to match all events. The condition fields must be in the following format:

<field> <operator> <value>

For example, the following snippet defines a filter that matches all POST requests:

filters:
- name: postfilter
  action: accept
  condition: cqhm MATCH POST

Filter Fields¶

The log fields have already been discussed in the Formats section above. For a reference to the available log field names, see Log Fields. Unlike with the log format specification, you do not wrap the log field names in any additional markup.

Filter Operators¶

The operators describe how to perform the matching in the filter rule, and may be any one of the following:

MATCH: True if the values of field and value are identical. Case-sensitive.
CASE_INSENSITIVE_MATCH: True if the values of field and value are identical. Case-insensitive.
CONTAIN: True if the value of field contains value (i.e. value is a substring of the contents of field). Case-sensitive.
CASE_INSENSITIVE_CONTAIN: True if the value of field contains value (i.e. value is a substring of the contents of field). Case-insensitive.
LT: True if the value of field is less than the value of value. Only valid for integer fields.
LTE: True if the value of field is less than or equal to the value of value. Only valid for integer fields.
GT: True if the value of field is greater than the value of value. Only valid for integer fields.
GTE: True if the value of field is less than or equal to the value of value. Only valid for integer fields.

Filter Values¶

The final component of a filter string specifies the value against which the name field will be compared.

For integer matches, all of the operators are effectively equivalent and require the field to be equal to the given integer. If you wish to match multiple integers, provide a comma separated list like this:

<field> <operator> 4,5,6,7

String matches work similarly to integer matches. Multiple matches are also supported via a comma separated list. For example:

<field> <operator> e1host,host2,hostz

For IP addresses, ranges may be specified by separating the first address and the last of the range with a single - dash, as 10.0.0.0-10.255.255.255 which gives the ranges for the 10/8 network. Other network notations are not supported at this time.

注釈

It may be tempting to attach multiple Filters to a log object reject multiple log fields (in lieu of providing a single comma separated list to a single Filter). Avoid this temptation and use a comma separated list of reject objects instead. Remember that you may not have multiple accept filter objects. Attaching multiple filters does the opposite of what you'd expect. If, for example, we had 2 accept log filters, each disjoint from the other, nothing will ever get logged on the given log object.

Logs¶

Up to this point, we've only described what events should be logged and what they should look like in the logging output. Now we define where those logs should be sent.

Three options currently exist for the type of logging output: ascii, binary, and ascii_pipe. Which type of logging output you choose depends largely on how you intend to process the logs with other tools, and a discussion of the merits of each is covered elsewhere, in Deciding Between ASCII or Binary Output.

The following subsections cover the attributes you should specify when creating your logging object. Only filename and format are required.

Name	Type	Description
filename	string	The name of the logfile relative to the default logging directory (set with `proxy.config.log.logfile_dir`). If this is set to `stdout` or `stderr`, then the stdout or stderr stream will be logged to, respectively.
format	string	a string with a valid named format specification.
header	string	If present, emitted as the first line of each new log file.
rolling_enabled	see below	Determines the type of log rolling to use (or whether to disable rolling). Overrides `proxy.config.log.rolling_enabled`.
rolling_interval_sec	number	Interval in seconds between log file rolling. Overrides `proxy.config.log.rolling_interval_sec`.
rolling_offset_hr	number	Specifies an hour (from 0 to 23) at which log rolling is guaranteed to align. Only has an effect if RollingIntervalSec is set to greater than one hour. Overrides `proxy.config.log.rolling_offset_hr`.
rolling_size_mb	number	Size, in megabytes, at which log files are rolled.
rolling_min_count	number	Specifies the minimum number of rolled logs to keep.
filters	array of filters	The optional list of filter objects which restrict the individual events logged. The array may only contain one accept filter.

Enabling log rolling may be done globally in records.yaml, or on a per-log basis by passing appropriate values for the rolling_enabled key. The latter method may also be used to effect different rolling settings for individual logs. The numeric values that may be passed are the same as used by proxy.config.log.rolling_enabled. For convenience and readability, the following predefined variables may also be used in logging.yaml:

log.roll.none: Disable log rolling.
log.roll.time: Roll at a certain time frequency, specified by RollingIntervalSec and RollingOffsetHr.
log.roll.size: Roll when the size exceeds RollingSizeMb.
log.roll.both: Roll when either the specified rolling time is reached or the specified file size is reached.
log.roll.any: Roll the log file when the specified rolling time is reached if the size of the file equals or exceeds the specified size.

Examples¶

The following is an example of a format that collects information using three common fields:

formats:
- name: minimalfmt
  format: '%<chi> , %<cqu> , %<pssc>'

The following is an example of a format that uses aggregate operators to produce a summary log:

formats:
- name: summaryfmt
  format: '%<LAST(cqts)>:%<COUNT(*)>:%<SUM(psql)>'
  interval: 10

The following is an example of a filter that will cause only REFRESH_HIT events to be logged:

filters:
- name: refreshhitfilter
  action: accept
  condition: pssc MATCH REFRESH_HIT

The following is an example of a log specification that creates a local log file for the minimal format defined earlier. The log filename will be minimal.log because we select the ASCII logging format.

logs:
- mode: ascii
  filename: minimal
  format: minimalfmt

The following is an example of a log specification that creates a local log file using the summary format from earlier, and only includes events that matched the REFRESH_HIT filter we created.

logs:
- mode: ascii
  filename: refreshhit_summary
  format: summaryfmt
  filters:
  - refreshhitfilter