airbnb / plog

84%

master: 64%

LAST BUILD ON BRANCH pierre/datadog/plog4_support

branch: pierre/datadog/plog4_support

CHANGE BRANCH

x

Reset

• pierre/datadog/plog4_support
• davide/simpler_fragment_name
• master
• pierre/actual_tags
• pierre/defrag/fix_leak
• pierre/doc/build_upload
• pierre/feature/52
• pierre/fix/defragmenter_warning
• pierre/fix_loglog_stats
• pierre/fragment_fixes
• pierre/gigantic
• pierre/guava_services
• pierre/http
• pierre/kafka/cleanup_stats_reporting
• pierre/minor/0
• pierre/revamp_stats
• pierre/stress/initial
• pierre/style_cleanup_20140512
• pierre/tag_tests
• pierre/test/all_the_things
• pierre/test_eop_logger
• pierre/tests/fragmented_message
• pierre/v2.0
• pierre/v3.0.0
• pierre/v4alpha
• pierre/versions_without_v
• typo
• v2.0
• v2.0.2
• v4.0.0-BETA

datadog: initial plog4 support work

- The plugin is now easy to test with:

        $ PYTHONPATH=~/.datadog-agent/agent ./plog.py

- The logic is now in the order keys are returned by plog.
- Our helper names follow the datadog naming convention.
- The Kafka logic is gone as Kafka is merely a handler now.
- Handlers are fully supported, with the following design:
  - If the same handler appears multiple times in the pipeline,
    we sum-merge stats from all instances.
    Note that handler instances could return different names to
    avoid this, if it ever really makes sense, as naming is dynamic.
  - Nested dictionaries are supported by compiling paths in the form
    'key1.key2.key3.'[...]
  - As we walk the tree, as soon as we see a non-dictionary, we switch
    to "nested summing", so [1,[2,{a:3,b:'c'}] = 1+2+3 = 6.