versipellis

Versipellis Configuration Reference

[collector.http] Sub-Section

url - server address, in the form "http[s]://host[:port][/path][?query]"

• Required
• If the host is an IPv6 address, it must be enclosed in square brackets, e.g., [fe80::1]
• Query parameters may be specified here as a suffix (e.g., ?param1=value1&param2=value2), but it’s recommended to specify them in query (see the notes there)

method - HTTP request method, a.k.a. verb, for retrieving data

• Optional
• Default: "GET"
• Options (case insensitive): "GET", "PATCH", "POST", "PUT"

query - HTTP query parameters

• Optional
• Query parameters may be specified as a suffix in url, but it’s recommended to specify them here:
  • Values in query override values in url, if they have the same parameter names
  • Values in query are allowed to contain dynamic expressions which are evaluated during runtime (although this is not implemented yet), while url must be static

• The TOML file format supports multiple representation options for key-value pairs:

  [collector.http]
  # ...
  query = { param1 = "value1", param2 = "value2" }
  

  [collector.http]
  # ...
  query = {
      param1 = "value1",
      param2 = "value2",
  }
  

  [collector.http]
  # ...
  query.param1 = "value1"
  query.param2 = "value2"
  

  [collector.http]
  # ...
  
  [collector.http.query]
  param1 = "value1"
  param2 = "value2"
  

headers - HTTP request headers

• Optional

• The TOML file format does not allow multiple instances of the same field, so if you want to specify a header with multiple values, specify them as a single comma-separated string:

  accept = "text/plain, application/json, application/xml"
  

• The TOML file format supports multiple representation options for key-value pairs:

  [collector.http]
  # ...
  headers = { header1 = "value1", header2 = "value2" }
  

  [collector.http]
  # ...
  headers = {
      header1 = "value1",
      header2 = "value2",
  }
  

  [collector.http]
  # ...
  headers.header1 = "value1"
  headers.header2 = "value2"
  

  [collector.http]
  # ...
  
  [collector.http.headers]
  header1 = "value1"
  header2 = "value2"
  

body or body_file - HTTP request body

• Optional
• Not allowed when method = "GET"
• Use at most one of them, not both (they’re mutually exclusive):
  • Inline - usually when the content is small and simple
  • Relative or absolute path to a file (e.g., "config/body.json" or "/path/body.xml") - when it’s large, complex, or sensitive

max_body_size - maximum byte size of a server response’s whole body

• Optional
• Default: 10485760 (10,485,760 bytes = 10 MiB)
• Valid range: any positive integer up to 104857600 (100 MiB)
• 0 and negative integers are normalized to the default
• Integers greater than the maximum are normalized to the maximum

max_headers_size - maximum byte size of a server response’s headers (as a whole, not each key-value pair)

• Optional
• Default: 1048576 (1,048,576 bytes = 1 MiB)
• Valid range: any positive integer up to 104857600 (100 MiB)
• 0 and negative integers are normalized to the default
• Integers greater than the maximum are normalized to the maximum

timeout - maximum amount of time for each HTTP client request to complete

• Optional
• Default: "5s" (5 seconds)
• Format: string containing decimal numbers, each with a unit suffix, e.g., m (minutes), and s (seconds)
• Special case: "0" and negative values (e.g., "-1s") = no client-side timeout

format - discard all HTTP response headers, pass only the HTTP response body

• Optional
• Default: none (don’t discard HTTP response headers)
• Options:
  • "json" (an array if the body contains multiple objects, otherwise an object)
  • "json_array" (always an array, even when there are fewer than 2 objects)
  • "ndjson" (NDJSON, a.k.a. JSONL)
  • "raw" (opaque byte payload)
• This format is a downstream preference…
  • It does not have to be identical to the body’s own content type, for example: you can receive JSON and pass it as NDJSON, and vice versa
  • Either way, the configured sender may override this
• If the body can’t be parsed in the configured format, the entire HTTP response is moved to the Dead-Letter Queue (DLQ)

[collector.http.tls] Sub-Section

• Optional for HTTP/1.1, required for HTTP/2
• See this page

[collector.http.retries] Sub-Section

• Optional
• See this page

This site is open source. Improve this page.