Skip to content

Latest commit

 

History

History
438 lines (319 loc) · 35.7 KB

configuration.md

File metadata and controls

438 lines (319 loc) · 35.7 KB

Configuration

The entire configuration is contained in config.yml.

The following are common parameters which it is possible to specify in the configuration.

Parameter Description Expected format Example Required
notificationIntervalSeconds Defines the amount of seconds after which an alert can be repeated. An alert is repeated only if the event that triggered it is not yet solved. Please, don't set this value to Infinity, use instead alertOnlyOnce. An integer 1800 Yes
monitoredPrefixesFiles The list of files containing the prefixes to monitor. See here for more informations. A list of strings (valid .yml files) -prefixes.yml Yes
logging A dictionary of parameters containing the configuration for the file logging. Yes
logging.directory The directory where the log files will be generated. The directory will be created if not existent. A string logs Yes
logging.logRotatePattern A pattern with date placeholders indicating the name of the file. This pattern will also indicate when a log file is rotated. A string with date placeholders (YYYY, MM, DD, ss, hh) YYYY-MM-DD Yes
logging.compressOnRotation Indicates if when a file gets rotates it has to be compressed or not. A boolean true Yes
logging.maxFileSizeMB Indicates the maximum file size in MB allowed before to be rotated. This allows to rotate files when logRotatePattern still the same but the file is too big An integer 15 Yes
logging.maxRetainedFiles Indicates the maximum amount of log files retained. When this threshold is passed, files are deleted. An integer 10 Yes
logging.useUTC If set to true, logs will be reported in UTC time. Is set to false or missing, the timezone of the machine will be used. This parameter affects only the timestamp reported at the beginning of each log entry, it doesn't affect the time reported in the data/alerts which is always in UTC. A boolean true No
checkForUpdatesAtBoot Indicates if at each booth the application should check for updates. If an update is available, a notification will be sent to the default group. If you restart the process often (e.g., debugging, experimenting etc.) set this to false to avoid notifications. Anyway, BGPalerter checks for updates every 10 days. A boolean true Yes
processMonitors A list of modules allowing various ways to check for the status of BGPalerter (e.g., API, heartbeat). See here for more information. No
httpProxy Defines the HTTP/HTTPS proxy server to be used by BGPalerter and its submodules (reporters/connectors/monitors). See here for more information. A string http://usr:psw@ prxy.org:8080 No
persistStatus If set to true, when BGPalerter is restarted the list of alerts already sent is recovered. This avoids duplicated alerts. The process must be able to write on disc inside .cache/. A boolean true No
generatePrefixListEveryDays This parameter allows to automatically re-generate the prefix list after the specified amount of days. Set to 0 to disable it. It works only if you have one prefix list file and if you have used BGPalerter to automatically generate the file (and not if you edited prefixes.yml manually). An integer 0 No
rpki A dictionary containing the RPKI configuration (see here for more details). Yes
groupsFile A file containing user groups definition (see here for more details). A string groups.yml No
rest A dictionary containing the parameters to run the server for all APIs provided by BGPalerter. No
rest.host The IP/host on which the APIs will be reachable. The default value is localhost, this means the API will not be reachable from another host. To make it public use null or 0.0.0.0. A string or null localhost No
rest.port The port of the REST API. The default value is 8011. An integer 8011 No

The following are advanced parameters, please don't touch them if you are not doing research/experiments.

Parameter Description Expected format Example Required
environment You can specify various environments. The values "production" (not verbose) and "development" (verbose) will affect the verbosity of the error/debug logs. The value "research" is explained here. Other values don't affect the functionalities, they will be used to identify from which environment the log is coming from. A string production Yes
alertOnlyOnce A boolean that, if set to true, will prevent repetitions of the same alert in the future (which it doesn't make sense for production purposes). In this case notificationIntervalSeconds will be ignored. If set to true, the signature of all alerts will be cached in order to recognize if they already happened in the past. This may lead to a memory leak if the amount of alerts is considerable. A boolean false No
pidFile A file where the PID of the BGP alerter main process is recorded. A string bgpalerter.pid No
maxMessagesPerSecond A cap to the BGP messages received, over such cap the messages will be dropped. The default value is way above any production rate. Changing this value may be useful only for research measurements on the entire address space. An integer 6000 No
multiProcess If set to true, the processing of the BGP messages will be distributed on two processes. This may be useful for research measurements on the entire address space. It is discouraged to set this to true for normal production monitoring. A boolean false No
fadeOffSeconds If an alert is generated but cannot be yet squashed (e.g., not reached yet the thresholdMinPeers), it is inserted in a temporary list which is garbage collected after the amount of seconds expressed in fadeOffSeconds. Due to BGP propagation times, values below 5 minutes can result in false negatives. An integer 360 No
checkFadeOffGroupsSeconds Amount of seconds after which the process checks for fading off alerts. An integer 30 No
volume Defines a directory that will contain all the files used by BGPalerter. See here before modifying this. A string /home/bgpalerter/ No

Composition

You can compose the tool with 3 main components: connectors, monitors, and reports.

  • Connectors retrieve/listen to the data from different sources and transform them to a common format.
  • Monitors analyze the data flow and produce alerts. Different monitors try to detect different issues.
  • Reports send/store the alerts, e.g., by email or to a file. Reports can also provide the data triggering such alerts.

In config.yml.example there are all the possible components declarations (similar to the one of the example below). You can enable the various components by uncommenting the related block.

Example of composition:

connectors:
  - file: connectorRIS
    name: ris
    
monitors:
  - file: monitorHijack
    channel: hijack
    name: basic-hijack-detection

  - file: monitorPath
    channel: path
    name: path-matching
    params:
      thresholdMinPeers: 10

reports:
  - file: reportFile
    channels:
      - hijack
      - path
    params:
      persistAlertData: false
      alertDataDirectory: alertdata/

Each monitor declaration is composed of:

Attribute Description
file Name of the file containing the monitor implementation. Monitor implementations are in the monitors directory.
channel The name of the channel that will be used by the monitor to dispatch messages. If the inserted name doesn't correspond to an already existent channel, a new channel is created.
name The name associated to the monitor. Multiple monitors with the same implementation can be loaded with different names. This name will be used to annotate messages in order track from where they are coming from.
params A dictionary of parameters that can be useful for the functioning of the monitor. Different monitors with the same implementation can be initialized with different parameters.

Each report declaration is composed of:

Attribute Description
file Name of the file containing the report implementation. Report implementations are in the reports directory.
params A dictionary of parameters that can be useful for the functioning of the report. It is common to have group declarations among the parameters. Different reports with the same implementation can be initialized with different parameters.
channels A list of channels the report module will listen (never write). Different reports with the same implementation can be initialized with a different list of channels to listen.

Each connector is composed of:

Attribute Description
file Name of the file containing the connector implementation. Report implementations are in the connectors directory.
params A dictionary of parameters that can be useful for the functioning of the connector. E.g. the data source url, password, socket options
name The name that will be used to identify this connector and to annotate logs and messages.

Connectors will always send the BGP updates to all the channels. The BGP updates have all the same format.

Connectors

Connectors retrieve/listen to the data from different sources and transform them to a common format.

Possible connectors are:

connectorRIS

It connects to RIPE RIS https://ris-live.ripe.net/ and receives BGP updates coming from 600+ peers.

Parameters for this connector module:

Parameter Description
url WebSocket end-point of RIS, which currently is ws://ris-live.ripe.net/v1/ws/
subscription Dictionary containing the parameters required by RIS. Refer to the official documentation for details.
noProxy If there is a global proxy configuration (see here), this parameter if set to true allows the single module to bypass the proxy.
carefulSubscription If this parameter is set to true (default), the RIS server will stream only the data related to our prefix. This is an advanced parameter useful only for research purposes.
perMessageDeflate Enable gzip compression on the connection.
disableCanary ConnectorRIS automatically receives BGP announcements about RIS beacons. RIS beacons are prefixes periodically announced and withdrawn for research purposes. BGPalerter uses these beacons to detect faulty data streams. By setting this parameter to true, you will disable such a check.
blacklistSources A list of prefixes, IPs, and AS numbers of blacklisted collectors' peers. Data coming from these collectors' peers will be ignored.

connectorRISDump

It connects to the RIPEstat's BGPlay API and retrieves a RIS dump about the monitored resources. The retrieved dump is 2 hours old, due to limitations on the API side.

Without this connector, when you start BGPalerter the monitoring will start based on new BGP updates. This means that you will not receive alerts before a new BGP update is propagated; e.g., if one of your prefixes is already hijacked when you start BGPalerter, you will not get notified immediately.

This connector runs only in the two following conditions:

  • you enable it in the config.yml (commented out by default);
  • you didn't start BGPalerter in the last two hours.

connectorTest

Connector used for testing purposes, it provokes all types of alerting. Needed to run the tests (npm run test) .

Monitors

Monitors analyze the data flow and produce alerts. Different monitors try to detect different issues.

Possible monitors are:

monitorHijack

This monitor has the logic to detect basic hijack. In particular, it will monitor for all the declared prefixes and will trigger an alert when:

  • The origin AS of the prefix is different from the ones specified in the configuration.
  • A more specific of the prefix has been announced by an AS which is different from the ones specified.
  • The BGP update declares an AS_SET as origin and at least one of the AS in the AS_SET is not specified in the configuration.

Example of alert:

The prefix 2a00:5884::/32 (description associated with the prefix) is announced by AS15563 instead of AS204092

Parameters for this monitor module:

Parameter Description
thresholdMinPeers Minimum number of peers that need to see the BGP update before to trigger an alert.
notificationIntervalSeconds It overwrite the global notificationIntervalSeconds for this specific monitor. See here the definition.
noProxy If there is a global proxy configuration (see here), this parameter if set to true allows the single module to bypass the proxy.
maxDataSamples Maximum number of collected BGP messages for each alert which doesn't reach yet the thresholdMinPeers. Default to 1000. As soon as the thresholdMinPeers is reached, the collected BGP messages are flushed, independently from the value of maxDataSamples.

monitorVisibility

This monitor has the logic to detect loss of visibility. In particular, it will monitor for all the declared prefixes and will trigger an alert when:

  • The prefix is not visible anymore from at least thresholdMinPeers peers.

Example of alert:

The prefix 165.254.225.0/24 (description associated with the prefix) has been withdrawn. It is no longer visible from 4 peers

Parameters for this monitor module:

Parameter Description
thresholdMinPeers Minimum number of peers that need to see the BGP update before to trigger an alert.
notificationIntervalSeconds It overwrite the global notificationIntervalSeconds for this specific monitor. See here the definition. Useful if you are monitoring a beacon.
noProxy If there is a global proxy configuration (see here), this parameter if set to true allows the single module to bypass the proxy.
maxDataSamples Maximum number of collected BGP messages for each alert which doesn't reach yet the thresholdMinPeers. Default to 1000. As soon as the thresholdMinPeers is reached, the collected BGP messages are flushed, independently from the value of maxDataSamples.

monitorPath

This monitor detects BGP updates containing AS_PATH which match particular regular expressions.

Example: The prefixes list of BGPalerter has an entry such as:

165.254.255.0/24:
  asn: 15562
  description: an example on path matching
  ignoreMorespecifics: false
  path:
    - match: ".*2194,1234$"
      notMatch: ".*5054.*"
      matchDescription: detected scrubbing center
    - match: ".*123$"
      notMatch: ".*5056.*"
      matchDescription: other match

Path is a list of matching rules, in this way multiple matching rules can be defined for the same prefix (rules are in OR).

More about path matching here.

An alert will be generated for example when a BGP announcements for 165.254.255.0/24 or a more specific contains an AS_PATH terminating in 2194,1234 but not containing 5054. The generated alert will report the matchDescription field.

Example of alert:

Matched "an example on path matching" on prefix 98.5.4.3/22 (including length violation) 1 times

Parameters for this monitor module:

Parameter Description
thresholdMinPeers Minimum number of peers that need to see the BGP update before to trigger an alert.
notificationIntervalSeconds It overwrite the global notificationIntervalSeconds for this specific monitor. See here the definition.
noProxy If there is a global proxy configuration (see here), this parameter if set to true allows the single module to bypass the proxy.
maxDataSamples Maximum number of collected BGP messages for each alert which doesn't reach yet the thresholdMinPeers. Default to 1000. As soon as the thresholdMinPeers is reached, the collected BGP messages are flushed, independently from the value of maxDataSamples.

monitorNewPrefix

This monitor has the logic to detect unexpected change of configuration in the form of new more specific prefixes announced by the correct AS.

In particular, it will monitor for all the declared prefixes and will trigger an alert when:

  • A sub-prefix of the monitored prefix starts to be announced by the same AS declared for the prefix.

Example: The prefixes list of BGPalerter has an entry such as:

50.82.0.0/20:
   asn: 58302
   description: an example
   ignoreMorespecifics: false

If in config.yml monitorNewPrefix is enabled you will receive alerts every time a more specific prefix (e.g., 50.82.4.0/24) is announced by AS58302.

Example of alert:

A new prefix 165.254.255.0/25 is announced by AS15562. It should be instead 165.254.255.0/24 (description associated with the prefix) announced by AS15562

Parameters for this monitor module:

Parameter Description
thresholdMinPeers Minimum number of peers that need to see the BGP update before to trigger an alert.
notificationIntervalSeconds It overwrite the global notificationIntervalSeconds for this specific monitor. See here the definition.
noProxy If there is a global proxy configuration (see here), this parameter if set to true allows the single module to bypass the proxy.
maxDataSamples Maximum number of collected BGP messages for each alert which doesn't reach yet the thresholdMinPeers. Default to 1000. As soon as the thresholdMinPeers is reached, the collected BGP messages are flushed, independently from the value of maxDataSamples.

monitorAS

This monitor will listen for all announcements produced by the monitored Autonomous Systems and will detect when a prefix, which is not in the monitored prefixes list, is announced. This is useful if you want to be alerted in case your AS starts announcing something you didn't intend to announce (e.g., misconfiguration, typo).

Example: The prefixes list of BGPalerter has an options.monitorASns list declared, such as:

50.82.0.0/20:
   asn: 58302
   description: an example
   ignoreMorespecifics: false

options:
 monitorASns:
   58302:
     group: default

If in config.yml monitorAS is enabled, you will receive alerts every time a prefix not already part of the prefixes list is announced by AS58302.

If AS58302 starts announcing 45.230.23.0/24 an alert will be triggered. This happens because such prefix is not already monitored (it's not a sub prefix of 50.82.0.0/20).

You can generate the options block in the prefixes list automatically. Refer to the options -s and -m of the auto-configuration.

Example of alert:

AS2914 is announcing 2.2.2.3/22 but this prefix is not in the configured list of announced prefixes

Parameters for this monitor module:

Parameter Description
thresholdMinPeers Minimum number of peers that need to see the BGP update before to trigger an alert.
notificationIntervalSeconds It overwrite the global notificationIntervalSeconds for this specific monitor. See here the definition.
noProxy If there is a global proxy configuration (see here), this parameter if set to true allows the single module to bypass the proxy.
maxDataSamples Maximum number of collected BGP messages for each alert which doesn't reach yet the thresholdMinPeers. Default to 1000. As soon as the thresholdMinPeers is reached, the collected BGP messages are flushed, independently from the value of maxDataSamples.
skipPrefixMatch It allows to send all the AS-related alerts to the AS monitoring rule instead of giving priority to prefix rules. Default false.

monitorRPKI

This monitor will listen for all announcements produced by the monitored Autonomous Systems and for all the announcements involving any of the monitored prefixes (independently of who is announcing them), and it will trigger an alert if any of the announcements is RPKI invalid or not covered by ROAs (optional).

This monitor is particularly useful when:

  • Before RPKI deployment, since it will let you test what announcements will be invalid after creating the ROAs.
  • During RPKI deployment, since it will let you know if any of your announcements are invalid.
  • After you deployed RPKI, in order to be sure all future BGP configurations will be covered by ROAs.

Example: The prefixes list of BGPalerter has the following entries:

103.21.244.0/24:
   asn: 13335
   description: an example
   ignoreMorespecifics: false

options:
 monitorASns:
   13335:
     group: default

If in config.yml monitorRPKI is enabled, you will receive alerts every time:

  • 103.21.244.0/24 is announced and it is not covered by ROAs, or the announcement is RPKI invalid;
  • AS13335 announces something that is not covered by ROAs, or the announcement is RPKI invalid;
  • A prefix you announce used to be covered by a ROA but such ROA is no longer available (e.g., RPKI repositories past failures: ARIN, RIPE NCC)

Examples of alerts:

The route 103.21.244.0/24 announced by AS13335 is not RPKI valid
The route 1.2.3.4/24 announced by AS1234 is not covered by a ROA
The route 1.2.3.4/24 announced by AS1234 is no longer covered by a ROA

You need to configure your RPKI data source as described here.

Parameters for this monitor module:

Parameter Description
checkUncovered If set to true, the monitor will alert also for prefixes not covered by ROAs in addition of RPKI invalid prefixes.
checkDisappearing If set to true, the monitor will check also for disappearing ROAs. Important: set this feature to false if you have monitorROAS enabled; monitorROAS provides diffs including disappearing ROAs.
thresholdMinPeers Minimum number of peers that need to see the BGP update before to trigger an alert.
notificationIntervalSeconds It overwrite the global notificationIntervalSeconds for this specific monitor. See here the definition.
noProxy If there is a global proxy configuration (see here), this parameter if set to true allows the single module to bypass the proxy.
maxDataSamples Maximum number of collected BGP messages for each alert which doesn't reach yet the thresholdMinPeers. Default to 1000. As soon as the thresholdMinPeers is reached, the collected BGP messages are flushed, independently from the value of maxDataSamples.
cacheValidPrefixesSeconds Amount of seconds ROAs get cached in order to identify RPKI repository malfunctions (e.g., disappearing ROAs). Default to 7 days.

monitorROAS

This monitor will periodically check the ROAs involving any of your ASes or prefixes. In particular, it will report about: ROAs involving your resources being edited, added or removed; expiring ROAs; TA malfunctions. You need to configure your RPKI data source as described here. Note, while BGPalerter will perform the check near real time, many RIRs have delayed ROAs publication times.

Example: The prefixes list of BGPalerter has the following entries:

1.2.3.4/24:
   asn: 1234
   description: an example
   ignoreMorespecifics: false
   group: noc1

options:
 monitorASns:
   2914:
     group: noc2

If in config.yml monitorROAS is enabled, you will receive alerts every time:

  • A ROA that is, or was, involving 1.2.3.4/24 is added/edited/removed (based on the prefix 1.2.3.4/24 matching rule).
  • A ROA that is, or was, involving AS2914 is added/edited/removed (based on the monitorASns section).

Important 1: for a complete monitoring, configure also the monitorASns section. Setting only prefix matching rules is not sufficient: prefix matching rules are based on the longest prefix match, less specific ROAs impacting the prefix will NOT be matched. On the other hand, setting only the monitorASns section is instead perfectly fine for ROA monitoring purposes.

Important 2: prefix matching rules have always priority on monitorASns rules. If an alert matches both a prefix rule and an AS rule, it will be sent only to the prefix rule, except if the checkOnlyASns params is set to true (see parameters below). In the example above, a ROA change impacting 1.2.3.4/24 is only sent to the user group noc1 and not to noc2. Other ROA changes impacting a prefix not in the list (no prefix matching rule) will be sent to noc2 instead.

Important 3: alerts about the generic health status of TAs are generated according to the provided VRP file. This types of alerts are not necessarily related to the monitored resources and they are send to the default user group.

Example of alerts:

ROAs change detected: removed <1.2.3.4/24, 1234, 25, apnic>; added <5.5.3.4/24, 1234, 25, apnic>

Possible TA malfunction or incomplete VRP file: 24% of the ROAs disappeared from APNIC

This monitor also alerts about ROAs expiration.

Note: This feature requires a vrp file having an expires field for each vrp, currently supported only by rpki-client. To enable this feature, provide a file generated with rpki-client (version 7.1 and newer) or use vrpProvider: rpkiclient in your rpki configuration (more info).

ROAs are affected by a series of expiration times:

  • Certificate Authority's "notAfter" date;
  • each CRL's "nextUpdate" date;
  • each manifest's EE certificate notAfter, and each manifests eContent "nextUpdate";
  • the ROA's own EE certificate "notAfter".

The field expire must be the closest expiration time of all of the above.

This means that you can receive an alert about an expiring ROA, where the expiration is not due to the ROA itself but due to some other expiring component of the validation chain.

Example of alerts:

The following ROAs will expire in less than 2 hours: <1.2.3.4/24, 1234, 25, apnic>; <5.5.3.4/24, 1234, 25, apnic>

Possible TA malfunction or incomplete VRP file: 24% of the ROAs are expiring in APNIC

Parameters for this monitor module:

Parameter Description
enableDiffAlerts Enables alerts showing edits impacting ROAs for the monitored resources. Default true
enableExpirationAlerts Enables alerts about expiring ROAs. Default true.
enableExpirationCheckTA Enables alerts about TA malfunctions detected when too many ROAs expire in the same TA. Default true.
enableDeletedCheckTA Enables alerts about TA malfunctions detected when too many ROAs are deleted in the same TA. Default true.
enableAdvancedRpkiStats Enables RPKI data enrichment on TA malfunction alerts and expiring ROAs alerts. It increases memory usage. Default true.
roaExpirationAlertHours If a ROA is expiring in less than this amount of hours, an alert will be triggered. The default is 2 hours. I strongly suggest to keep this value, ROAs are almost expiring every day, read above what this expiration time means.
checkOnlyASns If set to true (default), ROAs diff alerts will be generated based only on the ASns contained in the monitorASns of prefixes.yml. This means that no ROA diffs will be matched against prefix matching rules (see example above). If you are monitoring the origin AS of your prefixes, leave this option to true to avoid noise.
toleranceExpiredRoasTA The percentage of expiring ROAs in a single TA tolerated before triggering a TA malfunction alert. If you specify a numeric value, this will be used for all the TAs. Alternatively, you can specify a value for each TA (see config.yml.example).
toleranceDeletedRoasTA The percentage of deleted ROAs in a single TA tolerated before triggering a TA malfunction alert. If you specify a numeric value, this will be used for all the TAs. Alternatively, you can specify a value for each TA (see config.yml.example).

monitorPathNeighbors

The component monitorPathNeighbors allows to monitor for unexpected neighbor ASes in AS paths. The list of neighbors can be specified in prefixes.yml inside the monitorASns sections.

Refer to the documentation for this monitor.

Reports

Reports send/store the alerts, e.g., by email or to a file. Reports can also provide the data triggering such alerts. Refer to the report's documentation.