Advanced AnalyticsExabeam Advanced Analytics Administration Guide

Table of Contents

Troubleshoot Advanced Analytics Data Ingestion Issues

Hardware and Virtual Deployments Only

If events appear incorrectly in Advanced Analytics, track a log as it's ingested into Advanced Analytics to identify where and when it encounters any issues.

An event may appear incorrectly or not at all because there was an issue with ingesting data into Advanced Analytics:

To identify which of these issues caused your problem, track the logs as they're processed into the Advanced Analytics system. You can track logs based on criteria you specify, or specific logs.

If you track logs based on criteria you specify, you can only track the corresponding events as they're processed through the Analytics Engine. You create a JSON file that specifies which event fields the corresponding events must contain and the issue you're troubleshooting. Then, you run a Python script. The Python script adds a flag, EXA_ENABLE_TRACE_FILTERING, to all events with the event fields you specified. The flag marks the log so the Analytics Engine knows to record additional information about them. The script also sends the JSON file to the Analytics Engine for every node in your cluster.

If you track specific set of logs, you can track them as they're processed through both LIME and the Analytics Engine. You add a flag, EXA_ENABLE_TRACE, to the end of the log. This field marks the log so LIME and the Analytics Engine records additional information about it. Then, you ingest the log using the Syslog protocol.

As logs are ingested into Advanced Analytics, the Analytics Engine returns messages about what's happening to the logs; for example, if LIME parsed your log, which parser it used, and how it parsed the log. These messages are saved to a file called exabeam.log.

Use the returned messages to identify the problem and take steps to resolve it.

Returned Messages for Troubleshooting Data Ingestion Issues

Before you investigate your data ingestion issues, understand how the returned messages are structured and what they mean.

When troubleshooting potential data ingestion issues, you track logs as they're ingested through Log Ingestion Messaging Engine (LIME) or the Analytics Engine. After the log is ingested, the tracking mechanism returns messages to exabeam.log describing what happened as the log was ingested; for example:

| - | 2020-04-29 16:38:43.731 |   6.620 | INFO  | c.exabeam.lime.connector.SyslogImporter  | Created SyslogLine, objType=com.exabeam.lime.connector.SyslogImporter$SyslogLine, obj=SyslogLine(1585766883718,2020-04-01 18:48:03.718000+00:00,2020-04-01T14:48:03.718694-04:00,eamagnt01,"filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=joshfoxcroft@gmail.com suborg="" EXA_ENABLE_TRACE",None)
| - | 2020-04-29 16:38:45.530 |   8.419 | INFO  | c.exabeam.lime.event.LogToEventConverter | Created message info., objType=com.exabeam.lime.util.MessageInfo, obj=MessageInfo(exabeam_time=2020-04-01 18:48:03.718000+00:00 exabeam_reported_time=2020-04-01T14:48:03.718694-04:00 exabeam_host=eamagnt01 exabeam_raw=filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=joshfoxcroft@gmail.com suborg=" EXA_ENABLE_TRACE ,MessageType(proofpoint-m2,dlp-email-alert,yyyy-MM-dd HH:mm:ss,false,Vector(mod=session cmd=data rcpt=),ArrayBuffer(FieldExtractor(\sx=({xid}.+?)\s+(\w+=|$),ListBuffer('xid)), FieldExtractor(\srcpt=({recipient}[^@=]+?@({external_domain_recipient}.+?))\s+(\w+=|$),ListBuffer('recipient, 'external_domain_recipient))),None,None,Some(Proofpoint),Some(Proofpoint DLP)),msgType: "proofpoint-m2", dataType: "dlp-email-alert", msgOrder: "0", time: "1585766883718", rawLogTime: "1585766883718", timeFormat: "yyyy-MM-dd HH:mm:ss", fields: ('EXA_ENABLE_TRACE,"f._2"), ('recipient,"f._2"), ('external_domain_recipient,"f._2"), ('xid,"f._2"),true)
| - | 2020-04-29 16:38:45.559 |   8.449 | INFO  | com.exabeam.event.EventBuilderHelper     | Found handlers: Vector(com.exabeam.event.EventHandler@4388f8d3, com.exabeam.event.EventHandler@596c9550, com.exabeam.event.EventHandler@358d9f22, com.exabeam.event.EventHandler@6f4bf4f8), objType=com.exabeam.bar.message.Message, obj=msgType: "proofpoint-m2", dataType: "dlp-email-alert", msgOrder: "0", time: "1585766883718", rawLogTime: "1585766883718", timeFormat: "yyyy-MM-dd HH:mm:ss", fields: ('EXA_ENABLE_TRACE,"f._2"), ('recipient,"f._2"), ('external_domain_recipient,"f._2"), ('xid,"f._2")

Each output message has four parts:

<summary>, objType=<objType>, filteredBy=, <obj=<obj>
  • A short summary of what happened as the log was ingested

  • objType – Describes what object was traced: SyslogLine, MessageInfo, Message, Event.

    • SyslogLine – A raw log as it's ingested into LIME.

    • MessageInfo – A raw log that's been identified to an event type.

    • Message – A message that's been parsed from a raw log. Next, an Event Builder turns it into an event.Exabeam Parsers

    • Event – An event an Event Builder created. Information from the message has been mapped to event type fields.Exabeam Event Building

      If you're tracking multiple logs, the objType in your output messages is always Event because logs are tracked only as they're ingested through the Analytics Engine.

  • filteredBy – Appears only if you're tracking multiple logs; describes what models you're troubleshooting as indicated in the JSON file, under modelingFilter models.

  • obj – Describes what objType looks like at a given moment

For example, let's take this message:

| - | 2020-04-29 16:38:43.731 |   6.620 | INFO  | c.exabeam.lime.connector.SyslogImporter  | Created SyslogLine, objType=com.exabeam.lime.connector.SyslogImporter$SyslogLine, obj=SyslogLine(1585766883718,2020-04-01 18:48:03.718000+00:00,2020-04-01T14:48:03.718694-04:00,eamagnt01,"filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=joshfoxcroft@gmail.com suborg="" EXA_ENABLE_TRACE",None)
  • Created SyslogLine indicates that you're looking at the raw log.

  • The objType, objType=com.exabeam.lime.connector.SyslogImporter$SyslogLine, indicates that the tracking mechanism tracked the raw log.

  • The obj, obj=SyslogLine(1585766883718,2020-04-01 18:48:03.718000+00:00,2020-04-01T14:48:03.718694-04:00,eamagnt01,"filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=joshfoxcroft@gmail.com suborg="" EXA_ENABLE_TRACE",None), is what the raw log looked like.

Identify Which Event Details Were Enriched

Hardware and Virtual Deployments Only

Verify if an event is appearing incorrectly because the Analytics Engine enriched the event with incorrect information.

1. Identify logs to troubleshoot

If you don't know which log may be the issue, track a subset of logs that contain certain event fields. Create a JSON file that specifies these fields, then run a Python script to send the JSON file to Log Ingestion Messaging Engine (LIME) and the Analytics Engine for every node in your cluster.

Track specific logs

If you think that a specific set of logs may be the issue, track the log to verify this.

Add EXA_ENABLE_TRACE to the end of a raw log. EXA_ENABLE TRACE flags the log and lets the Analytics Engine know that it should log additional data. For example:

filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=barbara.salazar@acme.com suborg="" EXA_ENABLE_TRACE
Track logs based on event fields

If you don't know which log may be the issue, track a subset of logs that contain certain event fields. Create a JSON file that specifies these fields, then run a Python script to send the JSON file to Log Ingestion Messaging Engine (LIME) and the Analytics Engine for every node in your cluster.

  1. In a new JSON file, define the type of logs you're troubleshooting by event fields they contain. Create a key, eventFields.

    • To track a specific subset of logs, under eventFields, list the event fields these logs must contain as key/value pairs:

      {
          "eventFields":{
              event_type: ["eventtype1", "eventtype2"]
              source: ["source1", "source2"]
              user: ["user1", "user2"]
          },

      You can list any existing event field, including:

      • event_type

      • source

      • vendor

      • time

      • user

      • GetValue ('zone_info', src)

      To ensure that you cover an appropriate amount of logs, specify at least the event_type and user fields. For example, to track vpn-login and vpn-logout events for users Barbara Salazar and Sherri Lee, write:

      "eventFields" :{
          "user" : ["Blake Crouch", "Ted Chiang"]
          "event_type" : ["vpn-login", "vpn-logout"]
      },

      You can't use event field aliases that are used in MongoDB. When events are saved to MongoDB, the event fields are replaced with aliases; for example, the user event field is replaced with the g alias. In this JSON file, you must use user, not g.

    • To track all events, leave eventFields empty:

      {
          "eventFields":{},
  2. To enable the tracking mechanism, add a key, pipelineFilter, with the value "enabled:True":

    "pipelineFilter":{
        "enabled":true
    }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
    }
  3. Keys modelingFilter, ruleTriggeringFilter and eventDroppingFilter are used to troubleshoot other issues. You must add these keys and set value "enabled":false:

        "modelingFilter:{
        "enabled":false
        "models":[]
        },
        "eventDroppingFilter":{
            "enabled":false,
        },
        "ruleTriggeringFilter":{
            "enabled":false,
            "names":[]
        }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "modelingFilter":{
            "enabled":false
            "models":[]
        },
        "eventDroppingFilter":{
            "enabled":false
        },
        "ruleTriggeringFilter":{
            "enabled":false,
            "names":[]
        }
    }
  4. Add three fields: maxEventsToMark, maxMarkingTimeMs, and maxWorkingTime. To keep your system running correctly, these fields limit how many logs you can track and how long it takes to process them:

    "maxEventsToMark" : 50, 
    "maxMarkingTimeMs" : 3600000,
    "maxWorkingTimeMs" : 3600000
    • maxEventsToMark – The maximum number of logs to track.

    • maxMarkingTimeMs – How long, in seconds, ingested logs are flagged for tracking. After this period, the tracking mechanism may continue flagging logs for a few seconds and may flag more logs than maxEventsToMark allows. If you reach maxEventsToMark before maxMarkingTime, the tracking mechanism stops flagging logs.

    • maxWorkingTime – How long, in seconds, the tracking mechanism is enabled. After this period, the tracking mechanism may continue for an additional 30 seconds.

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "modelingFilter":{
            "enabled":false
        },
        "eventDroppingFilter":{
            "enabled":false
        },
        "ruleTriggeringFilter":{
        "enabled":false,
        "names":[]
        },
        "maxEventsToMark" : 50,
        "maxMarkingTimeMs" : 3600000,
        "maxWorkingTimeMs" : 3600000,
    }
  5. Save the file.

2. Ingest the logs

After you identify which logs to troubleshoot, ingest the log so the tracking mechanism returns data about how the log is being processed.

Ingest specific logs

If you're tracking a specific set of logs, ingest the log using the Syslog protocol. Ensure that you ingest one log at a time. If you ingest multiple logs at once, your system may run slower.

  • If you have command line access to your hardware appliance, ingest the log by appending to the Syslog file:

    cat <log file> >> /opt/exabeam/data/input/rsyslog/$(date '+%Y-%m-%dT%k').Internal.syslog.log
  • If you don't have command line access to your hardware appliance, ingest the log using a host that has a Syslog listener:

    cat <log file> | nc <appliance IP>
  • If Advanced Analytics fetches the log from another SIEM, send the raw log, with the added EXA_ENABLE_TRACE string, to that SIEM.

The log is tracked as it's ingested through LIME and the Analytics Engine.

Ingest logs based on criteria

If you're tracking logs based on criteria you specified, run a python script to ingest the log. This script adds a field, EXA_ENABLE TRACE_FILTERING, to all logs that match the event fields you specified in the JSON file. This field flags the logs and lets the Analytics Engine know that it should record additional data. It also sends the JSON file to LIME and the Analytics Engine for every node in your cluster.

  1. To start tracking the logs, run:

    ./martini_tracing_filter_control.py -c <yourfile.json>
  2. If you have a multi-node deployment and you want to aggregate and track logs from all nodes, run:

    run_shell_cmd_against_nodes.py -t

    To see what other commands you can run with run_shell_cmd_against_nodes.py, run:

    run_shell_cmd_against_nodes.py -h

    The log is tracked as it's ingested through the Analytics Engine only. It is not tracked as it's ingested through LIME.

3. Interpret Returned Messages

After the log is ingested, the tracking mechanism returns messages to exabeam.log describing what happened as the log was ingested. Use these messages to identify if LIME or the Analytics Engine encounters any errors when ingesting your log.

  1. Ensure that you understand what the returned messages mean.

  2. View the returned messages for the tracked logs:

    • To view all returned messages for all logs that were tracked, run:

      cat /opt/exabeam/data/logs/exabeam.log | grep "objType"=<objType>

    If no messages appear, it may be because:

    • One of the requirements in your JSON file is incorrect. Return to your JSON file and ensure that all fields in your JSON file are correct.

    • The event was discarded. Verify if this is the case.

  3. To understand what details the event was enriched with, compare the event before and after it enters the Analytics Engine:

    diff -u <(cat evt1 | sed 's/),(/),~(/g' | tr ',~' ',\n') <(cat evt2 | sed 's/),(/),~(/g' | tr ',~' ',\n') | grep -E "^\+"

    The command compares the log before it enters the Analytics Engine and before it enters the node. For example, the output may look like:

    (.env) [exabeam@cct210219-140012 ~]$ diff -u <(cat evt1 | sed 's/),(/),~(/g' | tr ',~' ',\n') <(cat evt2 | sed 's/),(/),~(/g' | tr ',~' ',\n') | grep -E "^\+"
    +++ /dev/fd/62	2021-02-19 14:57:39.581127012 +0000
    +| - | 2021-02-19 14:27:56.138 | 805.761 | INFO  | com.exabeam.martini.logging.LogTracing [MartiniContext-akka.actor.pipeline-dispatcher-44] | Received event at node='mainrun-context-enricher', objType=com.exabeam.martini.extractions.Event, obj=Event[time:2014-05-15 23:08:58,eType:vpn-login,source:VPN,vendor:Juniper VPN,id:2885@m,rawLogTime:1400206138000,optRawLogs:None,optRawLogRefs:None,optAlertUniqueId:None,sessionId:user63-20140515230858,sessionLabel:vpn-in,sessionStage:login,lockoutId:NA,lockoutStage:NA,fields:,(count(user, 'vpn-login'),1),
    +(duration,0),
    +(count(getvalue('isp', src_ip), 'app-login', 'app-activity', 'authentication-successful', 'vpn-login', 'remote-logon'),1),
    +(getvalue('country_code', src_ip),IL),
    +(distinctcount(badge_id, 'physical-access'),0),
    +(distinctcount(getvalue('zone_info', dest)),0),
    +(count(event_type, 'remote-access', 'remote-logon', 'local-logon', 'kerberos-logon', 'ntlm-logon', 'vpn-login', 'account-creation', 'account-deleted', 'member-added', 'member-removed', 'account-switch', 'app-login', 'app-activity'),1),
    +(count(event_type, 'remote-access', 'remote-logon', 'local-logon', 'kerberos-logon', 'ntlm-logon', 'vpn-login', 'account-password-reset', 'account-password-change', 'account-creation', 'account-deleted', 'member-added', 'member-removed', 'account-switch', 'app-login', 'app-activity', 'privileged-access', 'privileged-object-access', 'audit-policy-change', 'audit-log-clear', 'authentication-successful', 'database-login', 'nac-logon', 'physical-access', 'account-unlocked', 'account-unlocked-dup', 'account-enabled', 'account-disabled', 'member-added-dup', 'passwordvault-account-switch-dup', 'dlp-email-alert-out'),1),
    +(getvalue('isp', src_ip),Bezeq International),
    +(count(src_ip, 'vpn-login'),1),
    +(activity_types,Set(vpn)),
    +(count(getvalue('country_code', src_ip), 'app-login', 'app-activity', 'vpn-login', 'authentication-successful', 'remote-logon'),1),
    
  4. If the event wasn't enriched with details you expect, check your parsers or enrichment configuration for errors:

    • If the missing detail is in the raw log, verify if the parser is parsing your logs correctly.

    • If the detail is enriched from a context table, verify that your enrichment configuration and context management settings are correct.

Identify If an Event Was Discarded

Hardware and Virtual Deployments Only

Verify if an event is appearing incorrectly because the Analytics Engine incorrectly discarded the event.

First, create a JSON file that identifies the logs you're troubleshooting. Then, run a Python script that starts tracking logs as they're ingested into your system. The script saves any additional information logged during this process to exabeam.log.

1. Identify logs to troubleshoot

To troubleshoot an issue, you track logs as they're ingested into Advanced Analytics. If you think that one, specific log may be the issue, track just that log. If you're unsure about which log is causing the issue, track multiple logs based on criteria you specify, like event type, source, and user.

Track specific logs

If you think that a specific set of logs may be the issue, track the log to verify this.

Add EXA_ENABLE_TRACE to the end of a raw log. EXA_ENABLE TRACE flags the log and lets the Analytics Engine know that it should log additional data. For example:

filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=barbara.salazar@acme.com suborg="" EXA_ENABLE_TRACE
Track logs based on event fields

If you don't know which log may be the issue, track a subset of logs that contain certain event fields. Create a JSON file that specifies these fields, then run a Python script to send the JSON file to Log Ingestion Messaging Engine (LIME) and the Analytics Engine for every node in your cluster.

  1. In a new JSON file, define the type of logs you're troubleshooting by event fields they contain. Create a key, eventFields.

    • To track a specific subset of logs, under eventFields, list the event fields these logs must contain as key/value pairs:

      {
          "eventFields":{
              event_type: ["eventtype1", "eventtype2"]
              source: ["source1", "source2"]
              user: ["user1", "user2"]
          },

      You can list any existing event field, including:

      • event_type

      • source

      • vendor

      • time

      • user

      • GetValue ('zone_info', src)

      To ensure that you cover an appropriate amount of logs, specify at least the event_type and user fields. For example, to track vpn-login and vpn-logout events for users Barbara Salazar and Sherri Lee, write:

      "eventFields" :{
          "user" : ["Blake Crouch", "Ted Chiang"]
          "event_type" : ["vpn-login", "vpn-logout"]
      },

      You can't use event field aliases that are used in MongoDB. When events are saved to MongoDB, the event fields are replaced with aliases; for example, the user event field is replaced with the g alias. In this JSON file, you must use user, not g.

    • To track all events, leave eventFields empty:

      {
          "eventFields":{},
  2. To enable the tracking mechanism, add a key, pipelineFilter, with the value "enabled:True":

    "pipelineFilter":{
        "enabled":true
    }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
    }
  3. To track whether the specified events are going to be discarded, add a key, eventDroppingFilter with the value "enabled":true":

    "eventDroppingFilter":{
        "enabled":true
    }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "eventDroppingFilter":{
            "enabled":true
        },
    }
  4. Keys ruleTriggeringFilter and modelingFilter are used to troubleshoot other issues. You must add these keys and set value "enabled":false:

        "ruleTriggeringFilter":{
            "enabled":false,
            "names":[]
        },
        "modelingFilter":{
            "enabled":false,
            "models":[]
        }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "eventDroppingFilter":{
            "enabled":true
        },
        "ruleTriggeringFilter":{
            "enabled":false,
            "names":[]
        },
        "modelingFilter":{
            "enabled":false,
            "models":[]
        }
    }
  5. Add three fields: maxEventsToMark, maxMarkingTimeMs, and maxWorkingTime. To keep your system running correctly, these fields limit how many logs you can track and how long it takes to process them:

    "maxEventsToMark" : 50, 
    "maxMarkingTimeMs" : 3600000,
    "maxWorkingTimeMs" : 3600000
    • maxEventsToMark – The maximum number of logs to track.

    • maxMarkingTimeMs – How long, in seconds, ingested logs are flagged for tracking. After this period, the tracking mechanism may continue flagging logs for a few seconds and may flag more logs than maxEventsToMark allows. If you reach maxEventsToMark before maxMarkingTime, the tracking mechanism stops flagging logs.

    • maxWorkingTime – How long, in seconds, the tracking mechanism is enabled. After this period, the tracking mechanism may continue for an additional 30 seconds.

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "eventDroppingFilter":{
            "enabled":true
        },
        "ruleTriggeringFilter":{
        "enabled":false,
        "names":[]
        }
        "modelingFilter":{
            "enabled":false,
            "models":[]
        },
        "maxEventsToMark" : 50,
        "maxMarkingTimeMs" : 3600000,
        "maxWorkingTimeMs" : 3600000,
    }
  6. Save the file.

2. Ingest the logs

After you identify which logs to troubleshoot, ingest the log so the tracking mechanism returns data about how the log is being processed.

Ingest specific logs

If you're tracking a specific set of logs, ingest the log using the Syslog protocol. Ensure that you ingest one log at a time. If you ingest multiple logs at once, your system may run slower.

  • If you don't have command line access to your hardware appliance, ingest the log using a host that has a Syslog listener:

    cat <log file> | nc locahost 514
  • If you have command line access to your hardware appliance, ingest the log by appending to the Syslog file:

    cat <log file> >> /opt/exabeam/data/input/rsyslog/$(date '+%Y-%m-%dT%k').Internal.syslog.log
  • If Advanced Analytics fetches the log from another SIEM, send the raw log, with the added EXA_ENABLE_TRACE string, to that SIEM.

Ingest logs based on criteria

If you're tracking logs based on criteria you specified, run a python script to ingest the log. This script adds a field, EXA_ENABLE TRACE_FILTERING, to all logs that match the event fields you specified in the JSON file. This field flags the logs and lets the Analytics Engine know that it should record additional data. It also sends the JSON file to LIME and the Analytics Engine for every node in your cluster.

  1. To start tracking the logs, run:

    ./martini_tracing_filter_control.py -c <yourfile.json>
  2. If you have a multi-node deployment and you want to aggregate and track logs from all nodes, run:

    run_shell_cmd_against_nodes.py -t

    To see what other commands you can run with run_shell_cmd_against_nodes.py, run:

    run_shell_cmd_against_nodes.py -h

    The log is tracked as it's ingested through the Analytics Engine only. It is not tracked as it's ingested through LIME.

3. Interpret Returned Messages

After the log is ingested, the tracking mechanism returns messages to exabeam.log describing what happened as the log was ingested. Use these messages to identify if LIME or the Analytics Engine encounters any errors when ingesting your log.

  1. Ensure that you understand what the returned messages mean.

  2. To view all returned messages for all logs that were tracked, run:

    cat /opt/exabeam/data/logs/exabeam.log | grep "objType"=<objType>

    If no messages appear, it may be because:

    • One of the requirements in your JSON file is incorrect. Return to your JSON file and ensure that all fields in your JSON file are correct.

  3. To identify if an event was discarded, run:

    cat exabeam.log | grep -i 'discard .* due to'

    If the event was discarded, you may see messages indicating that the event was discarded and why, including:

    • discard share-access due to filter, objType=com.exabeam.martini.extractions.Event, obj=Event[...] – The event was discarded because of conditions specified in the DiscardIf hook in /opt/exabeam/config/custom/custom_exabeam_config.conf.

    • discard EVENT_TYPE due to login-filter, ... – The Analytics Engine couldn't find a session.

    • discard EVENT_TYPE due to time-before-login, ... – The event happened before the session.

    • discard EVENT_TYPE due to MaxUsers, ...MaxUsers ignores new user's events once the number of existing users reaches a specified threshold. The event was discarded because it met the MaxUsers threshold and the user is not in Active Directory.

    • discard EVENT_TYPE due to DropEventCondition-$CONDITION, ...Data Sanity detects misparsed or other erroneous data, then discards an event. The event was discarded because it matches one of the Data Sanity conditions in /opt/exabeam/config/custom/custom_exabeam_config.conf.

    • discard EVENT_TYPE due to MaxEvents – The event was discarded because the event type was disabled for that user, the user or asset was disabled, or if this is a user-related event, the user was neither logged in nor logged out of the session.

    • discard EVENT_TYPE due to default-filter, ... – The event was discarded because this is a computer logon event and asset sequences are disabled, this is a domain controller remote access event and the user field is a computer name ending with $, or the event was duplicated when both the domain controller and host sent the same log.

    • discard EVENT_TYPE due to error, ... – The event was discarded because of an exception in the Analytics Engine.

  4. If the event was discarded and this isn't what you expect, modify the DiscardIf hook in /opt/exabeam/config/custom/custom_exabeam_config.conf. Modifying the DiscardIf hook may have significant consequences for your system, so it's best that you contact Exabeam Customer Success to assist you.

    If you don't see these messages, the event wasn't discarded. Investigate what else may be causing your data ingestion issues.

Identify if a Model is Training

Hardware and Virtual Deployments Only

Verify if an event is appearing incorrectly because a model is training incorrectly on your events.

First, create a JSON file that identifies the logs and models you're troubleshooting. Then, run a Python script that starts tracking logs as they're ingested into your system. The script saves any additional information recorded during this process to exabeam.log. Use the returned messages in exabeam.log to investigate why your events might not be appearing correctly.

1. Identify logs to troubleshoot

To troubleshoot an issue, you track logs as they're ingested into Advanced Analytics. If you think that one, specific log may be the issue, track just that log. If you're unsure about which log is causing the issue, track multiple logs based on criteria you specify, like event type, source, and user.

Track specific logs

If you think that a specific set of logs may be the issue, track the log to verify this.

Add EXA_ENABLE_TRACE to the end of a raw log. EXA_ENABLE TRACE flags the log and lets the Analytics Engine know that it should log additional data. For example:

filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=barbara.salazar@acme.com suborg="" EXA_ENABLE_TRACE
Track logs based on event fields

If you don't know which log may be the issue, track a subset of logs that contain certain event fields. Create a JSON file that specifies these fields, then run a Python script to send the JSON file to Log Ingestion Messaging Engine (LIME) and the Analytics Engine for every node in your cluster.

  1. In a new JSON file, define the type of logs you're troubleshooting by event fields they contain. Create a key, eventFields.

    • To track a specific subset of logs, under eventFields, list the event fields these logs must contain as key/value pairs:

      {
          "eventFields":{
              event_type: ["eventtype1", "eventtype2"]
              source: ["source1", "source2"]
              user: ["user1", "user2"]
          },

      You can list any existing event field, including:

      • event_type

      • source

      • vendor

      • time

      • user

      • GetValue ('zone_info', src)

      To ensure that you cover an appropriate amount of logs, specify at least the event_type and user fields. For example, to track vpn-login and vpn-logout events for users Barbara Salazar and Sherri Lee, write:

      "eventFields" :{
          "user" : ["Blake Crouch", "Ted Chiang"]
          "event_type" : ["vpn-login", "vpn-logout"]
      },

      You can't use event field aliases that are used in MongoDB. When events are saved to MongoDB, the event fields are replaced with aliases; for example, the user event field is replaced with the g alias. In this JSON file, you must use user, not g.

    • To track all events, leave eventFields empty:

      {
          "eventFields":{},
  2. To enable the tracking mechanism, add a key, pipelineFilter, with the value "enabled:True":

    "pipelineFilter":{
        "enabled":true
    }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
    }
  3. To track whether a model is learning, add a key, modelingFilter with the value "enabled":true":

    "modelingFilter:{
        "enabled":true
    }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "modelingFilter":{
            "enabled":true
        },
    }
  4. Under the modelingFilter key, add another value, "models":[ ]. In the list, enter the models you're troubleshooting as a dictionary, with "name" as the key and the model as the value:

    "modelingFilter:{
        "enabled":true
        "models":[]
    }

    For example, to troubleshoot model UA-UI:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "modelingFilter":{
            "enabled":true
            "models":[{
                "name":"UA-UI"
            }]
        },
    }
  5. Keys ruleTriggeringFilter and eventDroppingFilter are used to troubleshoot other issues. You must add these keys and set value "enabled":false:

        "ruleTriggeringFilter":{
        "enabled":false,
        "names":[]
        },
        "eventDroppingFilter":{
            "enabled":false,
        }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "modelingFilter":{
            "enabled":true
            "models":[{
                "name":"UA-UI"
            }]
        },
        "eventDroppingFilter":{
            "enabled":false
        },
        "ruleTriggeringFilter":{
            "enabled":false,
            "names":[]
        }
    }
  6. Add three fields: maxEventsToMark, maxMarkingTimeMs, and maxWorkingTime. To keep your system running correctly, these fields limit how many logs you can track and how long it takes to process them:

    "maxEventsToMark" : 50, 
    "maxMarkingTimeMs" : 3600000,
    "maxWorkingTimeMs" : 3600000
    • maxEventsToMark – The maximum number of logs to track.

    • maxMarkingTimeMs – How long, in seconds, ingested logs are flagged for tracking. After this period, the tracking mechanism may continue flagging logs for a few seconds and may flag more logs than maxEventsToMark allows. If you reach maxEventsToMark before maxMarkingTime, the tracking mechanism stops flagging logs.

    • maxWorkingTime – How long, in seconds, the tracking mechanism is enabled. After this period, the tracking mechanism may continue for an additional 30 seconds.

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "modelingFilter":{
            "enabled":true
            "models":[{
                "name":"UA-UI"
            }]
        },
        "eventDroppingFilter":{
            "enabled":false
        },
        "ruleTriggeringFilter":{
        "enabled":false,
        "names":[]
        },
        "maxEventsToMark" : 50,
        "maxMarkingTimeMs" : 3600000,
        "maxWorkingTimeMs" : 3600000,
    }
  7. Save the file.

2. Ingest the logs

After you identify which logs to troubleshoot, ingest the log so the tracking mechanism returns data about how the log is being processed.

Ingest specific logs

If you're tracking a specific set of logs, ingest the log using the Syslog protocol. Ensure that you ingest one log at a time. If you ingest multiple logs at once, your system may run slower.

  • If you have command line access to your hardware appliance, ingest the log by appending to the Syslog file:

    cat <log file> >> /opt/exabeam/data/input/rsyslog/$(date '+%Y-%m-%dT%k').Internal.syslog.log
  • If you don't have command line access to your hardware appliance, ingest the log using a host that has a Syslog listener:

    cat <log file> | nc <appliance IP>
  • If Advanced Analytics fetches the log from another SIEM, send the raw log, with the added EXA_ENABLE_TRACE string, to that SIEM.

The log is tracked as it's ingested through LIME and the Analytics Engine.

Ingest logs based on criteria

If you're tracking logs based on criteria you specified, run a python script to ingest the log. This script adds a field, EXA_ENABLE TRACE_FILTERING, to all logs that match the event fields you specified in the JSON file. This field flags the logs and lets the Analytics Engine know that it should record additional data. It also sends the JSON file to LIME and the Analytics Engine for every node in your cluster.

  1. To start tracking the logs, run:

    ./martini_tracing_filter_control.py -c <yourfile.json>
  2. If you have a multi-node deployment and you want to aggregate and track logs from all nodes, run:

    run_shell_cmd_against_nodes.py -t

    To see what other commands you can run with run_shell_cmd_against_nodes.py, run:

    run_shell_cmd_against_nodes.py -h

    The log is tracked as it's ingested through the Analytics Engine only. It is not tracked as it's ingested through LIME.

3. Interpret Returned Messages

After the log is ingested, the tracking mechanism returns messages to exabeam.log describing what happened as the log was ingested. Use these messages to identify if LIME or the Analytics Engine encounters any errors when ingesting your log.

  1. Ensure that you understand what the returned messages mean.

  2. View the returned messages for the tracked logs:

    • To view all returned messages for all logs that were tracked, run:

      cat /opt/exabeam/data/logs/exabeam.log | grep "objType"=<objType>
    • To view all logs related to the models you specified under modelingFilter, run:

      cat /opt/exabeam/data/logs/exabeam.log | grep "filteredBy"=<model>

    If no messages appear, it may be because:

    • One of the requirements in your JSON file is incorrect. Return to your JSON file and ensure that all fields in your JSON file are correct.

    • The model isn't supposed to train on the event. Verify that the Analytics Engine didn't discard the event. If the Analytics Engine didn't discard the event, the event likely trained a different model than the one you're troubleshooting.

  3. Verify whether the model trained on the logs.

    For each returned message, you see different values for obj. If you see a message where obj is Going to update total histogram for model, the model trained on the event. Investigate what else may be causing your data ingestion issues.

  4. If you don't see any message where obj is Going to update total histogram for model, the model didn't train on the event. To investigate why this may be, work backwards from the latest returned message to fix a specific message. Some messages you may see include:

    • Might train model using event – This model is relevant to this event.

    • Going to early-train model – If the model RuleExpression contains num_observation=0, the Analytics Engine might train the model before triggering the associated rule.

    • Really modeling event_id= with model=trainIf=true means that the model should train on this event. firsttimeEvtTraining=true means that this is the first time the model is training on this event.

    • Skipping training model=... because of firstTimeEvtTraining=false and TrainIf=false, TrainifExpr=..., feature=...firstTimeEvtTraining=false means that the model previously tried training on this event. trainIf=false means that the model shouldn't train on this event.

    • Calculated fv/gfv for model: featureValues=, groupingFeatureValues= – Displays the featureValue and groupingFeatureValue attributes for the model, which may be helpful for debugging whether this model should train on these events.

    • Going to update total histogram for model, containerID=... – Model trained on the event. At the end of the session, results are written to local_model_db.

    • Won't do delayed training for event with containedId=... because skipTraining=true and isAfterWarmup=false – At this point, it's unclear which model, if any, is going to train on the event. If isAfterWarmup=false, no model will train on the event.

    If a returned message doesn't reflect behavior you expect, investigate other returned messages, your logs, or the model definition to discover if this is an error or if the model was never supposed to train on the event.

  5. If the model isn't training on the events and you think this is an error, verify if the event was enriched with the correct information.

Identify if a Rule is Triggering

Hardware and Virtual Deployments Only

Verify if an event is appearing incorrectly because a rule isn't triggering.

First, create a JSON file that identifies the logs and rules you're troubleshooting. Then, run a Python script that starts tracking logs as they're ingested into your system. The script saves any additional information recorded during this process to exabeam.log. Use the returned messages in exabeam.log to investigate why your events might not be appearing correctly.

1. Identify logs to troubleshoot

To troubleshoot an issue, you track logs as they're ingested into Advanced Analytics. If you think that one, specific log may be the issue, track just that log. If you're unsure about which log is causing the issue, track multiple logs based on criteria you specify, like event type, source, and user.

Track specific logs

If you think that a specific set of logs may be the issue, track the log to verify this.

Add EXA_ENABLE_TRACE to the end of a raw log. EXA_ENABLE TRACE flags the log and lets the Analytics Engine know that it should log additional data. For example:

filter_instance1[16174]: rprt s=3021wwe1wv m=1 x=3021wwe1wv-1 mod=session cmd=data rcpt=barbara.salazar@acme.com suborg="" EXA_ENABLE_TRACE
Track logs based on event fields

If you don't know which log may be the issue, track a subset of logs that contain certain event fields. Create a JSON file that specifies these fields, then run a Python script to send the JSON file to Log Ingestion Messaging Engine (LIME) and the Analytics Engine for every node in your cluster.

  1. In a new JSON file, define the type of logs you're troubleshooting by event fields they contain. Create a key, eventFields.

    • To track a specific subset of logs, under eventFields, list the event fields these logs must contain as key/value pairs:

      {
          "eventFields":{
              event_type: ["eventtype1", "eventtype2"]
              source: ["source1", "source2"]
              user: ["user1", "user2"]
          },

      You can list any existing event field, including:

      • event_type

      • source

      • vendor

      • time

      • user

      • GetValue ('zone_info', src)

      To ensure that you cover an appropriate amount of logs, specify at least the event_type and user fields. For example, to track vpn-login and vpn-logout events for users Barbara Salazar and Sherri Lee, write:

      "eventFields" :{
          "user" : ["Blake Crouch", "Ted Chiang"]
          "event_type" : ["vpn-login", "vpn-logout"]
      },

      You can't use event field aliases that are used in MongoDB. When events are saved to MongoDB, the event fields are replaced with aliases; for example, the user event field is replaced with the g alias. In this JSON file, you must use user, not g.

    • To track all events, leave eventFields empty:

      {
          "eventFields":{},
  2. To enable the tracking mechanism, add a key, pipelineFilter, with the value "enabled:True":

    "pipelineFilter":{
        "enabled":true
    }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
    }
  3. To track whether a rule is triggering, add a key, ruleTriggeringFilter with the value "enabled":true":

    "ruleTriggeringFilter":{
        "enabled":true,
    }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "ruleTriggeringFilter":{
            "enabled":true,
        },
    }
  4. Under the ruleTriggeringFilter key, add another value, "names":[ ]. In the list, enter the names of the rules you're troubleshooting in a list

    "ruleTriggeringFilter:{
        "enabled":true
        "names":[]
    }

    For example, to troubleshoot rules NEW-USER-F and UA-UC-Two:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "ruleTriggeringFilter":{
            "enabled":true,
            "names":["NEW-USER-F", "UA-UC-Two"]
        },
    }
  5. Keys modelingFilter and eventDroppingFilter are used to troubleshoot other issues. You must add these keys and set value "enabled":false:

        "modelingFilter":{
            "enabled":false,
            "models":[]
        },
        "eventDroppingFilter":{
            "enabled":false,
        }

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "ruleTriggeringFilter":{
            "enabled":true,
            "names":["NEW-USER-F", "UA-UC-Two"]
        },
        "modelingFilter":{
            "enabled":false
        },
        "eventDroppingFilter":{
            "enabled":false
        },
    }
  6. Add three fields: maxEventsToMark, maxMarkingTimeMs, and maxWorkingTime. To keep your system running correctly, these fields limit how many logs you can track and how long it takes to process them:

    "maxEventsToMark" : 50, 
    "maxMarkingTimeMs" : 3600000,
    "maxWorkingTimeMs" : 3600000
    • maxEventsToMark – The maximum number of logs to track.

    • maxMarkingTimeMs – How long, in seconds, ingested logs are flagged for tracking. After this period, the tracking mechanism may continue flagging logs for a few seconds and may flag more logs than maxEventsToMark allows. If you reach maxEventsToMark before maxMarkingTime, the tracking mechanism stops flagging logs.

    • maxWorkingTime – How long, in seconds, the tracking mechanism is enabled. After this period, the tracking mechanism may continue for an additional 30 seconds.

    For example:

    {
        "eventFields":{
            event_type: ["eventtype1", "eventtype2"]
            source: ["source1", "source2"]
            user: ["user1", "user2"]
        },
        "pipelineFilter":{
            "enabled":true
        },
        "ruleTriggeringFilter":{
            "enabled":true,
            "names":["NEW-USER-F", "UA-UC-Two"]
        },
        "modelingFilter":{
            "enabled":false
        },
        "eventDroppingFilter":{
            "enabled":false
        },
        "maxEventsToMark" : 50,
        "maxMarkingTimeMs" : 3600000,
        "maxWorkingTimeMs" : 3600000,
    }
  7. Save the file.

2. Ingest the logs

After you identify which logs to troubleshoot, ingest the log so the tracking mechanism returns data about how the log is being processed.

Ingest specific logs

If you're tracking a specific set of logs, ingest the log using the Syslog protocol. Ensure that you ingest one log at a time. If you ingest multiple logs at once, your system may run slower.

  • If you have command line access to your hardware appliance, ingest the log by appending to the Syslog file:

    cat <log file> >> /opt/exabeam/data/input/rsyslog/$(date '+%Y-%m-%dT%k').Internal.syslog.log
  • If you don't have command line access to your hardware appliance, ingest the log using a host that has a Syslog listener:

    cat <log file> | nc <appliance IP>
  • If Advanced Analytics fetches the log from another SIEM, send the raw log, with the added EXA_ENABLE_TRACE string, to that SIEM.

The log is tracked as it's ingested through LIME and the Analytics Engine.

Ingest logs based on criteria

If you're tracking logs based on criteria you specified, run a python script to ingest the log. This script adds a field, EXA_ENABLE TRACE_FILTERING, to all logs that match the event fields you specified in the JSON file. This field flags the logs and lets the Analytics Engine know that it should record additional data. It also sends the JSON file to LIME and the Analytics Engine for every node in your cluster.

  1. To start tracking the logs, run:

    ./martini_tracing_filter_control.py -c <yourfile.json>
  2. If you have a multi-node deployment and you want to aggregate and track logs from all nodes, run:

    run_shell_cmd_against_nodes.py -t

    To see what other commands you can run with run_shell_cmd_against_nodes.py, run:

    run_shell_cmd_against_nodes.py -h

    The log is tracked as it's ingested through the Analytics Engine only. It is not tracked as it's ingested through LIME.

3. Interpret Returned Messages

After the log is ingested, the tracking mechanism returns messages to exabeam.log describing what happened as the log was ingested. Use these messages to identify if LIME or the Analytics Engine encounters any errors when ingesting your log.

  1. Ensure that you understand what the returned messages mean.

  2. View the returned messages for the tracked logs:

    • To view all returned messages for all logs that were tracked, run:

      cat /opt/exabeam/data/logs/exabeam.log | grep "objType"=<objType>
    • To view all logs related to the rules you specified under ruleTriggeringFilter, run:

      cat /opt/exabeam/data/logs/exabeam.log | grep "filteredBy"=<rule>

    If no messages appear, it may be because:

    • One of the requirements in your JSON file is incorrect. Return to your JSON file and ensure that all fields in your JSON file are correct.

    • The event isn't supposed to trigger this rule. Ensure that the associated model trained on the event. If the associated model didn't train on the event, the event likely triggered another rule than the one you're troubleshooting.

    Look at rule definition, expression, events should be triggering n

    Check event and view log; something may not be referenced correctly in log; might indicate that log isn't parsed correctly

  3. Verify whether the rule triggered on the event.

    For each returned message, you see different values for obj. If you see a message where obj is ruleTriggered=true, the rule triggered on the event. Investigate what else may be causing your data ingestion issues.

  4. If you see a returned message where obj is ruleTriggered=false, the rule didn't train on the event and the RuleExpression was evaluated to false. To investigate, work backwards from the latest returned message to understand why this happened. Some messages include:

    • Going to evaluate – The rule was evaluated against the event, and your system will run further checks.

    • Rule was not evaluated because matchesRuleEventType=true and ignoreRule=true – The rule was ignored and not evaluated because ClassifyIf was evaluated to false or the event didn't match the rule type.

    • Computed ruleEventSeq=Seq(...) for the event. Going to evaluate rule for each entity – For asset-related rules, the rule will be evaluated against each asset.

    • Evaluating rule – The rule was evaluated against the event and passed all pre-checks. This may appear multiple times with new data; for example, as the rule evaluates against each asset.

    • Chained rule triggered or Chained rule was not triggered – A chained rule is a rule that depends on another rule. The chained rule's RuleExpression was evaluated to true or false, respectively.

    • Evaluated rule. Triggered=false – For asset-related rules, the rule was evaluated against each asset.

    • Skipping classification because of ClassifyIf=false, rule=...ClassifyIf indicates how often the rule should trigger. The rule wasn't evaluated because ClassifyIf was evaluated to false.

    • Skipping classification because of unknown model for rule=... – The rule wasn't evaluated because the rule or model wasn't configured correctly.

    • Rule was not evaluated because matchesRuleEventType=false and ignoreRule=trueClassifyIf indicates how often the rule should trigger. The rule was not triggered because ClassifyIf=false, or the rule or model wasn't configured correctly.

    If a returned message doesn't reflect behavior you expect, investigate other returned messages; or, compare your logs, the Advanced Analytics event, or the rule definition and expression to understand if this is an error or if the rule was never supposed to trigger on the event.

  5. If the rule isn't triggering on the events and you think this is an error, verify if the event was parsed correctly, or verify if the event was enriched with the correct information.