Logstash

- misc: clean up some excess debugging output
- feature: for tcp input, allow 'data_timeout => -1' to mean "never time out"

- deprecated: The redis 'name' and 'queue' options for both input and output
are now deprecated. They will be removed in a future version.
- feature: The redis input and output now supports both lists and channels.
- feature: Refactor runner to allow you to run multiple things in a single
process. You can end each instance with '--' flag. For example, to run one
agent and one web instance:
% java -jar logstash-blah.jar agent -f myconfig -- web
- feature: Add 'embedded' option to the elasticsearch output:
elasticsearch { embedded => true }
Default is false. If true, logstash will run an elasticsearch server
in the same process as logstash. This is really useful if you are just
starting out or only need one one elasticsearch server.
- feature: Added a logstash web backend feature for elasticsearch that tells
logstash to use the 'local' (in process) elasticsearch:
--backend elasticsearch:///?local
- feature: Added 'named_captures_only' option to grok filter. This will have
logstash only keep the captures you give names to - for example %{NUMBER}
won't be kept, but %{NUMBER:bytes} will be.
- feature: Add 'bind_host' option to elasticsearch output. This lets you choose the
address ElasticSearch client uses to bind to - useful if you have a
multihomed server.
- feature: The mongodb output now supports authentication
- bugfix: Fix bug in GELF output that caused the gelf short_message to be set as an
array if it came from a grok value. The short_message field should only
now be a string properly.
- bugfix: Fix bug in grep filter that would drop/cancel events if you had
more than one event type flowing through filters and didn't have a grep
filter defined for each type.
- misc: Updated gem dependencies (tests still pass)
- misc: With the above two points, you can now run a single logstash process
that includes elasticsearch server, logstash agent, and logstash web.

- Fix tcp input bug (LOGSTASH-88) that would drop connections.
- Grok patterns_dir (filter config) and --grok-patterns-dir (cmdline opt)
are now working.
- GELF output now properly sends extra fields from the log event (prefixed
with a "_") and sets timestamp to seconds-since-epoch (millisecond
precision and time zone information is lost, but this is the format GELF
asks for).
- Inputs support specifying the format of input data (see "format" and
"message_format" input config parameters).
- Grok filter no longer incorrectly tags _grokparsefailure when more than
one grok filter is enabled (for multiple types) or when an event has
no grok configuration for it's type.
- Fix bug where an invalid HTTP Referer: would break grok parsing of the
log line (used to expect %{URI}). Since Referer: is not sanitized in
the HTTP layer, we cannot assume it will be a well formed %{URI}.

- Fix crash bug caused by refactoring that left 'break' calls in code
that no longer used loops.

- Remove beanstalk support because the library (beanstalk-client) is GPL3. I
am not a lawyer, but I'm not waiting around to have someone complain about
license incompatibilities.
- fix bug in jar build

- logstash 'web' now allows you to specify the elasticsearch clustername;
--backend elasticsearch://[host[:port]]/[clustername]
- GELF output now supports dynamic strings for level and facility
https://logstash.jira.com/browse/LOGSTASH-83
- 'amqp' output supports persistent messages over AMQP, now. Tunable.
https://logstash.jira.com/browse/LOGSTASH-81
- Redis input and output are now supported. (Contributed by dokipen)
- Add shutdown processing. Shutdown starts when all inputs finish (like
stdin) The sequence progresses using the same pipeline as the
inputs/filters/outputs, so all in-flight events should finish getting
processed before the final shutdown event makes it's way to the outputs.
- Add retries to unhandled input exceptions (LOGSTASH-84)