Skip to content

Prometheus's expression language data types

<expression> := <metric_name>{<label_selectors>} <operator> <aggregation/functions>

Instant vector

  • A set of time series containing a single sample (datapoints) for each time series, evaluated at the a evaluation time ("now" if omitted)
  • The engine looks backward from the evaluation time and returns that series' most recent sample within the lookback window (default 5m)
services_http_requests_total{service="ronaldo",path="/api/version"}

{
  "metric": {
    "__name__": "services_http_requests_total",
    "aws_region": "us-east-2",
    "environment": "prod",
    "kubernetes_pod_name": "ronaldo-5975f7c77d-7kktn",
    "method": "get",
    "path": "/api/version",
    "service": "ronaldo",
    "status": "200",
  },
  "value": [
    1783977321, "1576"
  ]
},

Range vector

  • A set of time series containing a set of samples (datapoints) for each time series over a period of time
  • Range vectors exists so that functions can consume a window of raw data and these functions need a set of raw samples to operate on

  • The real job of range vectors is almost always to be immediately wrapped in a function (rate(), increase(), avg_over_time(), sum_over_time())

  • This is different from a range query (query_range API): the range vector syntax has no step concept and returns every raw stored sample in the window with its original scrape timestamp. query_range instead re-evaluates at each step grid point, keeping only the sample nearest each step (silently dropping others that fall between two steps) and stamping the result with the step timestamp, not the real sample timestamp.

  • When a bare range vector expression is sent to query_range directly, vicmetrics will automatically wrap it in a default_rollup()

rate(services_http_requests_total{service="ronaldo",path="/api/version"}[5m])

{
  "metric": {
    "__name__": "services_http_requests_total",
    "aws_region": "us-east-2",
    "environment": "prod",
    "kubernetes_pod_name": "prod-global-blue-ronaldo-deployment-5975f7c77d-7kktn",
    "method": "get",
    "path": "/api/version",
    "service": "ronaldo",
    "status": "200",
  },
  "values": [
    [ 1783978236, "1762" ],
    [ 1783978251, "1768" ],
    [ 1783978266, "1768" ],
    [ 1783978281, "1774" ],
    [ 1783978296, "1774" ],
    [ 1783978311, "1780" ],
    [ 1783978326, "1780" ],
    [ 1783978341, "1786" ],
    [ 1783978356, "1786" ],
    [ 1783978371, "1792" ],
    [ 1783978386, "1792" ],
    [ 1783978401, "1798" ],
    [ 1783978416, "1798" ],
    [ 1783978431, "1804" ],
    [ 1783978446, "1804" ],
    [ 1783978461, "1810" ],
    [ 1783978476, "1810" ],
    [ 1783978491, "1816" ],
    [ 1783978506, "1816" ],
    [ 1783978521, "1816" ],
    [ 1783978536, "1816" ]
  ]
},

ms # milliseconds
s # seconds
m # minutes
h # hours
d # days - assuming a day has always 24h
w # weeks - assuming a week has always 7d
y # years - assuming a year has always 365d

1h30m

Scalar

  • Float literals
  • A single floating point number (no labels)
23
-2.43
3.4e-9
0x8f
-Inf
NaN

String

  • String literal
  • Used in limited string-processing functions like label_replace()
  • Defined with quotes, double quotes or backticks
"this is a string"
'these are unescaped: \n \\ \t'
`these are not unescaped: \n ' " \t`