package com.postmates.cernan

Get desktop application:
View/edit binary Protocol Buffers messages

'AggregationMethod' - an interpretation signal Cernan maintains quantile summaries for all Telemetry samples. Not all sinks are capable of interpreting summaries natively. Cernan allows the client to set preferred aggregations over the summaries for reporting to 'flat' sinks. Sinks are allows to ignore AggregationMethod at their convenience. Additionally, aggregation time windows may be configured per-sink and are not controllable through the protocol.

Used in: Telemetry

• SUM = 1

  SUM keeps a sum of samples. This is often interpreted as a per-window counter.

• SET = 2

  SET preserves the last sample set into the Telemetry per time window.

• SUMMARIZE = 3

  SUMMARIZE produces a quantile summary of the input samples per time window. This is the default behaviour.

• BIN = 4

  BIN produces a histogram summary of the input samples per time window. The user will specify the bins' upper bounds.

'LogLine' - a bit of unstructure text One of cernan's gigs is picking up logs from disk and transforming them in-flight, shipping them off. This structure allows you to ship lines directly via the native protocol without having to round-trip through disk first.

Used in: Payload

• optional string path = 1

  unique 'location' of the log line

• optional string value = 2

  the line itself

• map<string, string> metadata = 3

  associated key/value metadata

• optional int64 timestamp_ms = 4

  milliseconds since the Unix epoch

'Payload' - the top-level structure in each on-wire payload Payload is a container for repeated Telemetry and LogLines. There's not much more to it than that.

• repeated Telemetry points = 2

• repeated LogLine lines = 3

'Telemetry' - a numeric measure of a thing Cernan's slightly more complicated gig is its 'telemetry' subsystem. Telemetry is defined as a name and time associated collection of measurements. In the structure we refer to these measurements as 'samples'. The Telemetry structure makes is possible to associate multiple samples in a single millisecond time window. Cernan will build a quantile structure over these samples but you may further choose aggregation interpretations by setting AggregationMethod.

Used in: Payload

• optional string name = 1

  the unique name of the telemetry

• repeated double samples = 2

  telemetry samples present in timestamp_ms

• optional bool persisted = 3

  persist metric across time windows

• optional AggregationMethod method = 4

  see below

• map<string, string> metadata = 5

  associated key/value metadata

• optional int64 timestamp_ms = 6

  milliseconds since the Unix epoch

• repeated double bin_bounds = 7

  BIN inclusive upper bounds