Monitors: how they work

Monitors are the core of the program. You can initialize their configuration with ripe-atlas-monitor init-monitor -m monitor_name: a monitor template file will be created and opened for customization using the preferred text editor (which can also be set within the global configuration file or via the $EDITOR environment variable).

How they work

You have a RIPE Atlas measurement;

  • probes involved in the measurement collect some results (ping, traceroute, ...);
  • a ripe-atlas-monitor‘s monitor is executed;
  • results for the aforementioned measurement are downloaded and elaborated;
  • for each result, the probe’s information (ID, country, ASN) are matched against a set of rules;
  • if a matching condition is found, the result collected by that probe is matched against a set of results you expected from that probe;
  • actions (email, syslog, external programs) are performed on the basis of this process’s output.

All this is written in YAML files, one for each monitor you want to configure:

descr: Check network reachability
- descr: Probes from France via AS64496
  src_country: FR
  expected_results: ViaAS64496
  actions: EMailToNOC
- descr: RTT from AS64499 and AS64500 below 50ms
  - 64499
  - 64500
  expected_results: LowRTT
  actions: EMailToNOC
    upstream_as: 64496
    rtt: 50
    kind: email
    subject: "ripe-atlas-monitor: unexpected results"
measurement-id: 123456789

For the complete syntax of monitors’ configuration file please refer to Monitor configuration syntax.

Kinds of monitors

Depending on the measurement they are configured to use and which command is used to run them, monitors can be grouped into 3 categories:

  • one-off monitors are those used to process one-off measurements: they are executed using the --latest argument of ripe-atlas-monitor to download only the latest results, or they can be executed using both the --start and --stop command line argument in order to define a specific time frame within which results are downloaded;
  • continous monitors are used to continously process results for those measurements which have not been stopped yet: results are downloaded and processed once every x seconds, where x is the interval value of the measurement itself; when the --start argument of ripe-atlas-monitor is used, results are downloaded starting at that time, otherwise results are downloaded starting from the timestamp of the last processed result;
  • streaming monitors, which are those that use RIPE Atlas result streaming.

The type of monitor is not written anywhere, it’s derived from the commands used to run ripe-atlas-monitor. For example, the same monitor can be run using ripe-atlas-monitor run -m MonitorName --measurement-id 123456 --latest to process the latest results from the measurement ID 123456, but also using ripe-atlas-monitor daemonize -m MonitorName to continously process results from the measurement reported in the measurement-id attribute of its configuration file. It can be also run in streaming mode, by using the --stream command line argument (provided that the measurement is still running).

Expected results criteria

Expected results can be of various kinds, depending on the measurement’s type, and various criteria can be used to verify collected results.

Traceroute measurements can be used to monitor AS path toward a destination, ping measurements to test network reachability and performance, SSL measurements to be sure that the certificates received by a probe match the expected fingerprints and that TLS connections are not hijacked on their way, DNS measurements to verify host name resolution.

For the full list of implemented criteria please read Monitor configuration syntax.

Advanced use

Configuration syntax “tricks” and internal labels allow to describe complex scenarios.

Excluding probes from processing

A rule with no expected_results and the process_next attribute to its default value False (or missing) allows to stop further processing for those probes which match the rule’s criteria:

- descr: Do not process results for probe ID 123 and 456
  - 123
  - 456

Match all probes except those...

The reverse attribute of a rule, when set to True, allows to match all the probes which do not meet the given criteria:

- descr: All probes except those from AS64496
  src_as: 64496
  reverse: True

Actions execution

The when attribute of an action can be used to set when it has to be performed:

  • on_match, the action is performed when the collected result matches one of the expected values, or when the rule has no expected results at all;
  • on_mismatch, the action is performed when the collected result does not match the expected values;
  • always, well, the action is always performed, independently of results.

Internal labels

Actions can be used to attach internal labels to probes on the basis of rules and results processing. These labels can be subsequently used to match probes against specific rules.

- descr: Set 'VIP' (Very Important Probe) label to ID 123 and 456
  - 123
  - 456
  process_next: True
  actions: SetVIPLabel
- descr: Set 'VIP' label to Italian probes too
  src_country: IT
  process_next: True
  actions: SetVIPLabel
- descr: VIPs must have low RTT
  internal_labels: VIP
  expected_results: LowRTT
    when: always
    kind: label
    op: add
    label_name: VIP

Integration with ripe-atlas-tools (Magellan)

Magellan is the official command-line client for RIPE Atlas. It allows, moreover, to create new measurements from the command line. It can be used, for example, in an action to create one-off measurements from the probes which fail expectations.

    descr: Create new traceroute msm from the probe which missed expectations
    kind: run
    path: ripe-atlas
    - measure
    - traceroute
    - --target
    - --no-report
    - --from-probes
    - $ProbeID