Seismometer Message


v: 3
time: 1376261660
  name: uptime
      value: 3205629.35
v: 3
time: 1376261660
  cluster: MongoDB
  environment: devel
  name: availability
    value: not_running
    severity: error


This is a handy reference of Seismometer message format. Full specification is available at <>.

Each Seismometer message describes some measurement (event). The result of the measurement can be a metric or set of metrics (free+used disk space, CPU usage, round-trip-time to some host, etc.), state (a server is running or is shut down, RAID storage healthy or degraded, etc), or both.

What kind of thing was measured (monitored aspect) is described by aspect name (field It’s usually a string like "disk space", "RAID status", "database availability". Where it was measured is described by location structure (sometimes called dimensions). Location usually includes host, and for disk related events it probably would contain device name or mount point.

In the case of metrics that don’t make much sense on their own and need to be kept bundled, like free/used/total disk space, a message can carry a set of values.

Message structure

Seismometer message is a JSON-compatible hash structure and is expected to have several fields:

  • v – numeric schema version; this document describes version 3
  • time – unix timestamp of the event
  • location – hash that describes where the event happened; keys should match /^[a-zA-Z0-9_]+$/ regexp, values are arbitrary strings
  • event – hash structure that describes the event; required fields are name and either state, vset, or both
    • – name of the aspect the event is about; name is a string with no further restrictions
    • event.state – state description (hash), if the message is about state
    • event.vset – value set (hash), if the message carries metrics
    • event.threshold_kept – name (string) of the state to set if the message carries value set with thresholds and all thresholds are kept; should be of the same format as event.state.value and defaults to "ok"
    • event.comment – optional string containing human-readable description of the event
    • event.interval – optional number of seconds between two consecutive events


Hash structure describing state of the monitored object. This structure has following fields:

  • value – state of the monitored object; the value should match /^[a-zA-Z0-9_]+$/ regexp
  • severity – optional severity of the state: "expected", "warning", or "error"; if not specified, defaults to "expected"


Hash containing a set of values (metrics). Key in this hash corresponds to name of a value and should match /^[a-zA-Z0-9_]+$/ regexp. Value is a hash structure with following fields:

  • value – number (integer or float) or null if no value could be measured (e.g. round-trip-time for a host that is down)
  • unit – optional string to specify units of value field
  • type – measurement type: "direct" (regular value), "accumulative" (counter that accumulates values over time, like network traffic counters), or "differential" (counter that is reset after each read); if not provided, defaults to "direct"
  • threshold_low, threshold_high – optional lists of hash structures that describe thresholds; such a structure has three fields: {"name": <name>, "value": <value>, "severity": <severity>}
    • <value> – if the measured value is greater than this (threshold_high) or less than this (threshold_low), this threshold is exceeded
    • <name> – if the threshold is exceeded, this becomes event.state.value
    • <severity> – if the threshold is exceeded, this becomes event.state.severity

See also: event.threshold_kept.

An example message with thresholds:

v: 3
time: 1376261660
  name: ping
  threshold_kept: ok
      value: 12.3
      unit: ms
        - {value: 30.0, name: warn, severity: warning}
        - {value: 50.0, name: crit, severity: error}
      value: 0
      unit: "%"
        - {value: 20, name: warn, severity: warning}
        - {value: 70, name: crit, severity: error}


data generated by monitoring probe as a result of a measurement
thing about the monitored object that is being measured; aspect tells how to interpret a value or state carried by a message
identification of the monitored object the message is about
short description of the monitored object’s status
level of operator’s attention needed by monitored object’s status; status is either expected, indicates a warning, or is an error and requires action
value set
collection of metrics about monitored object; usually a message only carries one metric, if any, but some metrics don’t make sense on their own and need to be presented together (e.g. free/used/total disk space)

Table Of Contents

Previous topic

Various helper modules

Next topic

Internals of the tools

This Page