Kentik Proxy Agent

The following topics cover the use of Kentik's NetFlow proxy agent, kproxy, to enable encrypted local flow from your organization's routers and switches to Kentik:

• Proxy Agent Overview
• kproxy Requirements
• kproxy Setup
• kproxy Download and Install
• kproxy Command Line
• SNMP Configuration File
• kproxy Debugging
• kproxy Syslog Parsing

The Kentik NetFlow proxy agent is covered in the following topics:

• About the Proxy Agent
• Kentik Traffic Flows
• Connections and Behavior
• Example Proxy Deployment

Kentik provides a free NetFlow proxy agent called "kproxy" that enables data collected from network devices to be sent securely to Kentik. The machine running kproxy isn't actually handling traffic directly, but instead collects flow records (NetFlow v5/v9, IPFIX, and sFlow) and SNMP and encrypts it locally before forwarding it to Kentik. A single instance of the kproxy executable can redirect flow for multiple routers and switches. Multiple servers across the network can run kproxy to distribute traffic and load.

In addition to the collection and encryption of device data, kproxy also performs the following functions:

• Rate limiting and resampling to keep maximum flows per second (FPS) within applicable license limits (see About Licenses).
• Parsing syslog files into flow records for inclusion in Kentik Data Engine (KDE) alongside data from other sources (see kproxy Syslog Parsing).

The steps involved in downloading, installation, and configuration of kproxy are described in the following topics below:

• The step-by-step routing of flow and SNMP to kproxy is covered in kproxy Setup.
• Downloading kproxy from Kentik is described in kproxy Download and Install.
• Configuration of kproxy for flow encryption is described in kproxy Command Line.

Notes:
- sFlow data received by kproxy is parsed to ensure that only fields relevant to flow analytics are forwarded to the Kentik ingest layer. kproxy does not forward the entire sFlow packet to Kentik.
- For assistance with any aspect of the setup process, please contact Customer Support.

The diagram below shows an overview of traffic flows between the Kentik NetFlow proxy agent, installed in a customer backbone or IT facility, and Kentik:

The following points describe agent behavior and the connection between customer devices, the Kentik proxy agent, and Kentik:

• kproxy transports all traffic received from customer devices securely to Kentik.
• The Kentik Data Engine and the Kentik portal handle data forwarded via kproxy identically to data received directly from customer devices.
• kproxy connects to any customer devices sending NetFlow telemetry to Kentik, and also to Kentik itself to send the encrypted data as well as to receive configuration information.
• In order to send traffic to Kentik, kproxy will build, for each NetFlow-sending device, two HTTPS sessions, one for flow traffic and the other for SNMP. All traffic is sent to Kentik using such HTTPS sessions, with a Kentik flow ingest server certificate.
• SNMP is converted into JSON format and NetFlow/sFlow/IPFIX is converted to a Kentik proprietary binary format.

The following diagram illustrates typical deployment of the Kentik proxy agent.

The following resources must be available (at minimum) to support the use of kproxy running on a VM:

• RAM allocation of 1GB per 1000 flow records per second (FPS).
  Note: Actual requirements vary depending on how “bursty” the flow is.
• One CPU core per 5000 flow records per second (before rate-limiting).

When kproxy is used for non-flow data, the following equivalents may be applied to the above requirements:

• Syslog: 1 line = 1 flow record.
• Streaming Telemetry: 1 interface metric = 1 flow record.
  Note: Using Streaming Telemetry with kproxy is distinct from using Streaming Telemetry for Kentik NMS (see NMS via Streaming Telemetry).

Notes:
- kproxy must be deployed on a separate server from any kprobe host agents sending flow to Kentik via kproxy.
- When running kproxy on a VM, the CPU must be dedicated to the instance.

Kproxy will listen for inbound UDP netflow/sflow on its configured port (default 9995) and relay this flow securely to flow.kentik.com in TLS-encrypted format. Upon receipt of flow from a configured device, the proxy will attempt (by default) to connect back to the IP address sending flow via SNMP to enumerate basic device information such as interface metadata and CPU.

Note: The kproxy UDP port is 9995 unless specified with the command line argument -port (see kproxy CLI Reference).

When used for encryption, kproxy pulls information from the Kentik system to determine which routers it will talk to. The routing of flow and SNMP to kproxy is enabled with the following steps:

1. Create a device in the Kentik portal (Devices » Add Device; see Network Devices) for each of the routers from which you'd like to send flow:
   - Set a device's Name and Description.
   - Set the device type as "Router" (even though you are sending through the agent).
   - Set the Device IP as the IP of the router that the agent will see the flow being received from. You may enter multiple IPs, comma separated, if there is the possibility that flow may source from multiple IPs for the said router. Private IPs are acceptable.
   - Set Device SNMP IP to the router IP that the agent will poll for SNMP.
   - Set Flow Type to the type of flow that the router is configured to export from.
   - Set Sample Rate to the rate at which the router is set to sample.
   - Save (Add) the device.
2. Check the system clock and time zone settings on the host server where kproxy will be deployed, which must be accurate to within a minute for kproxy to function correctly. Kentik recommends that hosts running kproxy use Network Time Protocol (NTP).
3. Determine the IP that kproxy will bind to on the host in order to receive flow.
4. Determine the port that the host will accept flow on (i.e. where you will point your routers too).
5. Download and install kproxy (see kproxy Download and Install).
6. Run kproxy, specifying the arguments described in kproxy Command Line. You'll likely want to test it initially from the command line and then place it into one of your startup scripts so that it begins on boot. If not placed in the background, the agent will run in the foreground and adhere to standard kill/end signals (e.g. run "nohup kproxy +options &" if running from the command line and exiting shell).
7. Configure your routers and switches to send flow to the kproxy instance (see Router Configuration).

Download and installation of kproxy — which may be deployed via a Debian/Ubuntu package, an RPM package, a Docker image, or via systemd — is covered in the following topics:

• kproxy Docker Install
• kproxy Package Install
• Deploying kproxy via systemd

To install kproxy via Docker:

1. In the Kentik portal, go to the kproxy Agents page (Settings » Proxy Agents), then click the Add kproxy Agent button to open the Adding a Kproxy Agent dialog.
2. In the dialog, click the Docker card, then copy the value of the --api_email and --api_token arguments shown in step 2 of the instructions.
3. In terminal, pull down the latest Kproxy image from Kentik’s Docker hub repository:
   docker pull kentik/kproxy:latest
4. To comply with security best practices, create an environment variable named KENTIK_API_TOKEN to store the authentication token, which is the --api_token value that you copied in step 2 above.
5. Run the kproxy docker image (use the --api_email value from step 2 above):
   docker run --name kproxy -d --net=host kentik/kproxy --api_email api-email-value@kentik.com

Notes:
- Storing the token as an environment variable prevents it from being exposed via a process (ps) command. If you prefer not to use an environment variable for the token, include an --api_token argument in the docker run command shown in step 5 above, using the value from step 2 above: --api_token api-token-value
- For additional options, including arguments for specifying a different IP address and/or port, see kproxy CLI Arguments.

Linux packages for kproxy are available for both Debian/Ubuntu and RPM (CentOS/RHEL) installations. To download and install:

1. In the Kentik portal, go to the kproxy Agents page (Settings » Proxy Agents), then click the Add kproxy Agent button to open the Adding a Kproxy Agent dialog.
2. In the dialog, click the Debian/Ubuntu Package or RPM Package card, then copy the value of the --api_token and --api_email argument shown in step 3 of the instructions.
3. In terminal, download and install the kproxy repository:
   - Debian/Ubuntu:
   curl -s https://packagecloud.io/install/repositories/kentik/kproxy/script.deb.sh | sudo bash
   sudo apt-get install kproxy
   - RPM:
   curl -s https://packagecloud.io/install/repositories/kentik/kproxy/script.rpm.sh | sudo bash
   yum install kproxy
   Note: If the "kproxy" package is not found, your operating system may be unsupported. To try a manual installation, select a version from the repository at https://packagecloud.io/kentik/kproxy.
4. To comply with security best practices, create an environment variable named KENTIK_API_TOKEN to store the authentication token, which is the --api_token value that you copied in step 2 above.
5. Run kproxy (use the --api_email value from step 2 above):
   kproxy --api_email api-email-value@kentik.com

Notes:
- Storing the token as an environment variable prevents it from being exposed via a process (ps) command. If you prefer not to use an environment variable for the token, include an --api_token argument in the docker run command shown in step 5 above, using the value from step 2 above: --api_token api-token-value
- To run kproxy via systemd, refer either to Deploying kproxy via systemd or to the Debian/Ubuntu or RPM instructions in the Add kproxy Agent on the Kentik portal's kproxy Agents page.
- For additional options, including arguments for specifying a different IP address and/or port, see kproxy CLI Arguments.
- To upgrade a package to the current version of kproxy, see Upgrading an Existing kproxy.
- Once kproxy has been downloaded and installed, continue with the steps outlined in kproxy Setup.

Once installed, a kproxy package must be kept up to date to ensure correct performance. For best results, Kentik recommends using your OS's package manager to upgrade existing instances of kproxy automatically. If using the package manager is not possible, then you can follow the steps below to perform a manual upgrade.

1. Check the version of the existing kproxy instance.

• Debian/Ubuntu:

# Command
dpkg -l | grep kproxy
# Response
ii kproxy-latest-ubuntu-16.04 2.3 amd64 no description given

• Red Hat/Centos/Fedora:

# Command
rpm -qa | grep kproxy
# Response
kproxy-latest-rhel_7-1.0-1.x86_64
# Use that package name in this command:
rpm -qi kproxy-latest-rhel_7-1.0-1.x86_64
# Response
Name : kproxy-latest-rhel_7
Version : 3.3

2. Kill the running kproxy process (all OS versions; placeholders highlighted).

# Command
ps -ef | grep kproxy
# Response
root  9979  9895  0 13:59 pts/0  00:00:00 /usr/bin/kproxy -api_pass password_string -api_email username@domain.suffix
# Command
sudo killall kproxy

3. Remove the existing kproxy package.

• Debian/Ubuntu

# Command
dpkg -l | grep kproxy
# Response
ii kproxy-latest-ubuntu-16.04 2.3 amd64 no description given
# Command
sudo dpkg -r kproxy-latest-ubuntu-16.04
# Response
(Reading database... 60028 files and directories currently installed.)
Removing kproxy-latest-ubuntu-16.04 (2.3)

• Red Hat/Centos/Fedora

# Command
rpm -qa | grep kproxy
# Response
kproxy-latest-rhel_7-1.0-1.x86_64
# Command
sudo rpm -e kproxy-latest-rhel_7-1.0-1.x86_64

4. Download and install the latest version of kproxy as described in kproxy Download and Install.
5. Check the version of the newly installed kproxy instance, and compare it with the version from step 1 to confirm successful installation.
6. Restart kproxy with the same command options returned in step 2 in response to the ps command. See step 6 of kproxy Setup.

To enable faster, more efficient setup of kproxy, especially when you need to deploy many instances, the kproxy installation process includes a kproxy service definition that can be run with Linux's systemd system and service manager:

• To use kproxy for flow encryption, run the default version of the service definition (replace placeholders with actual authentication credentials).
• To take advantage of other kproxy features, also edit the service definition to specify the desired kproxy behavior using the arguments covered in kproxy CLI Arguments.

The kproxy install process results in installation of two systemd-related files:

• /lib/systemd/system/kproxy.service
• /etc/default/kentik.env.sample

To run the service definition:

1. Copy the config file at /etc/default/kentik.env.sample to /etc/default/kentik.env
2. In the new /etc/default/kentik.env service definition file, replace the following highlighted placeholders with actual values associated with a Kentik-registered user:
   - KENTIK_API_TOKEN=XXX_FILL_ME_IN: The API token from the user's Authentication Settings.
   - KENTIK_API_EMAIL=foo@example.com: The user's email address.
3. To use this kproxy instance for tasks other than flow encryption, you can make additional edits in the service definition file, specifying the desired behavior using the arguments in kproxy CLI Arguments.
4. Choose the Linux command corresponding to how you want kproxy to start:
   - To start manually, use systemctl start kproxy
   - To start upon boot, use systemctl enable kproxy
5. To get logs (confirm that kproxy is running) and show the service state of the proxy, run the following Linux commands:
   journalctl -u kproxy
   systemctl status kproxy -l

The use of the kproxy command line is covered in the following topics:

• kproxy CLI Arguments
• Alternate DNS Lookup
• kproxy for Streaming Telemetry
• kproxy CLI Reference

The following topics cover the most common command line arguments for kproxy. For a brief reference to additional arguments, see kproxy CLI Reference.

Notes:
- If your organization is registered with Kentik in the EU you must set the -region argument to eu.
- If kproxy fails to launch, add the -verbose flag and try again so that you can provide the output to support@kentik.com in order to facilitate troubleshooting.
- Either " " or "=" can be used between an argument's name and value.
- Use -h to return a list of arguments.

The following arguments are required to authenticate access to kproxy:

• -api_email (required): The email address of a registered user as displayed on that user's API System page, which is accessed via the API button for that user on the Users page.
• -host (required): Set to one of the following interface IPs:
  - The IP of a single interface for kproxy to listen on;
  - 0.0.0.0 to listen on all interfaces.
• -region (required for EU customers): If your organization is registered with Kentik in the EU (rather than the US), you must include this argument and set the value to eu.

The command line arguments below are used when configuring kproxy as a NetFlow proxy agent.

• -port (optional): Set the port to listen on. If omitted, kproxy defaults to listening on port 9995.
• -dns (optional): The address, expressed as IP:Port (e.g. 127.0.0.1:5353), of an alternate (private) DNS server that is to be used for reverse DNS lookups instead of Kentik's default server. For usage, see Alternate DNS Lookup.
• -site_id (optional): The ID of a Kentik-registered site (see Manage Sites). If a different site ID is specified for two or more instances of kproxy, then a device sending flow to Kentik via one such instance will be able to use the same Sending IP address (see Device General Settings) as a device sending flow via another instance (see IP Overloading).
• -log_level (optional): Sets the level of logging. Valid values are debug, info (default), warn, error.
• -proxy_host (optional): Sets the IP that kproxy will use for child processes. The kprobe parent process spawns a child process to handle flow from each individual combination of device and time-slice; default is 127.0.0.1.
• -healthcheck (optional): The IP to use for the healthcheck service; default is 127.0.0.1.
• -syslog_config (optional): Specifies the path to a JSON configuration file that defines a schema which determines how KDE will, at ingest, parse the flow data contained in any syslog that is sent to this instance of kproxy (see kproxy Syslog Parsing).
• -bootstrap_devices (optional): A comma-separated list of device ids (e.g. "12002,1221") for which this instance of kproxy should initiate a child process without first waiting to receive flow. Used in scenarios such as using kproxy on a given device for SNMP only (no flow collection).
• -tee_kproxy (optional): Reserved for Kentik internal use.
• -sflow-agent-address (optional): Set the agent address as the sending IP rather than the original packet source IP address. Use this argument when sFlow may originate from multiple IP addresses on a single device or may traverse a NAT.

If you choose not to store the api_token as an environment variable (step 2 above), despite the risk of exposure via a process (ps) command, your command line will need to include the token using the following additional parameter:

• -api_token: A Kentik-generated string that kproxy will use to authenticate a registered user (must be the same user as for -api_email).

If an alternate DNS server is designated in the kproxy command line then at ingest the IP addresses in the flow records from this kproxy are looked up on the server at the provided IP:Port combination (e.g. 127.0.0.1:5353). The returned host names will be stored in Kentik's KDE backend in Source Hostname and Destination Hostname dimensions (see IP Info Dimensions. These directional hostname dimensions can be used for group-by and filtering in queries.

To avoid including public IP addresses from the Internet in the IP-to-hostname resolution, the designation of an alternate DNS server involves specifying to kproxy which IPs should be considered as "internal." The following command line arguments are used:

• -dns (optional): The IP address and the port of the local DNS that will be used for IP-to-hostname resolution. If an internal flag is included in the argument then the RFC1918 IP space and other common private ranges will be included in the IP addresses on which kproxy will perform IP-to-hostname resolution:
  - With flag: internal:ip-address:port
  - Without flag: ip-address:port
  Note: If the internal flag is not included then the -my_network argument is required.
• -my_network: The "internal" prefixes on which to perform IP-to-hostname resolution. The IPs to use may be specified as one of the following:
  - prefixes: A comma-separated list of IP prefixes.
  - netclass: A value that references the IPs listed in the Internal IPs field on the Kentik portal's Network Classification Page.
  - netclass, prefixes: The netclass value included in the list of prefixes.

Notes:
- If the -my_network argument is not provided but the -dns argument includes the internal flag then kproxy will perform IP-to-hostname resolution on only the RFC1918 IP space and other common private ranges.
- If the -my_network argument includes netclass and IPs are not specified in Network Classification then kproxy will fail with an error.

The table below shows how kproxy determines the IPs for IP-to-hostname resolution based on the two arguments above.

The following example shows how to configure the proxy agent (version 7.43.0 or newer) to derive the Hostname dimensions using an alternate DNS server that resolves the prefixes specified with my_network:

/usr/bin/kproxy -local_proxy http://10.10.2.2:3128 -proxy-http 192.168.2.16:3182 -bootstrap_devices 104884,88726 -dns 192.168.1.2:53 -my_network 192.168.1.0/24,192.168.2.0/24 -api_email=kproxy-agent+xxxx@kentik.com -log_level=info -host=0.0.0.0 -proxy_host=127.0.0.1 -healthcheck=0.0.0.0 -port=9995

Notes:
- To generate directional hostname dimensions your organization should be proxying flows through kproxy version 7.43.0 or newer. To enable alternate DNS resolution with earlier versions of kproxy, please contact Customer Support.
- When the IP addresses of an ingested flow record can't be resolved the directional hostname dimensions in the corresponding KDE record will be left unspecified.
- The -dns argument is distinct from Custom DNS, which affects only the display of hostnames in the portal and specifies an alternate DNS server for flows from all sources rather than just from an individual instance of kproxy.
- If an alternate DNS server is specified with -dns, the hostname returned for a given IP address won't be checked again for 24 hours.
- Kentik looks up host names from alternate servers on a "best effort" basis; if the rate of flow records from a given device is high then it may not be possible to resolve all of the IP addresses in those flows to host names.

Note: Using Streaming Telemetry with kproxy is distinct from using Streaming Telemetry for Kentik NMS (see NMS via Streaming Telemetry).

The following kproxy arguments are specified only when kproxy will be used to receive streaming telemetry data from devices and forward it to KDE (see SNMP and Streaming Telemetry). Leave these arguments unspecified in all other scenarios:

• -st_dialout_listener (required for ST on Junos or Cisco): Set to auto to enable receipt by kproxy of streaming telemetry data collected and sent either by devices running Junos or by Cisco IOS-XRv 9000 routers.
• -st_udp_bind (required for ST on Junos): Specifies the IP address and port on which kproxy should listen for streaming telemetry data collected and sent as UDP by devices running Junos (version 18.4R2.7 or later).
• -st_tcp_bind (required for ST on Cisco): Specifies the IP address and port on which kproxy should listen for streaming telemetry data collected and sent as TCP by devices such as Cisco IOS-XRv 9000 routers (version 6.2.3 or later).

The combination of above arguments used to enable Streaming Telemetry with Kentik varies depending on the data source:

• To enable ST via kproxy from devices running Junos:
  - Always use st_dialout_listener, set to auto.
  - Also use st_udp_bind, typically set to 0.0.0.0:9555.
• To enable ST via kproxy from Cisco IOS-XRv 9000 routers:
  - Always use st_dialout_listener, set to auto.
  - Also use st_tcp_bind, typically set to 0.0.0.0:9555.
• To enable ST from both Junos devices and Cisco IOS-XRv 9000 routers to the same instance of kproxy:
  - Always use st_dialout_listener, set to auto.
  - Also use both st_tcp_bind and st_udp_bind, both typically set to 0.0.0.0:9555.

Notes: Setting the binding arguments above to the Kentik-recommended value of 0.0.0.0:9555 enables kproxy to listen on the specified port on all IPs on the server.

The table below shows the complete set of kproxy arguments, along with their descriptions, as returned from the -h and --help commands. Either " " or "=" can be used between an argument's name and value. Arguments of type “bool” don’t require a value to be specified.

By default, the SNMP configuration (SNMP IP and SNMP Community) for a given device that sends flow to Kentik (e.g. router; see Supported Device Types) is learned by kproxy from that device's settings in the Kentik portal (see Device SNMP Settings). There may be circumstances, however, in which it is necessary (e.g. for security compliance) not to specify SNMP settings for a given router in the portal. In this case it is possible instead to specify the settings through kproxy configuration, using the optional -snmp_file command line argument to direct kproxy to get that information from a local config file.

When SNMP is configured with an external file, the required SNMP parameters are set from the values in that file. These values are described in the following table:

The required settings are stored in the config file as JSON key/value pairs. The following example shows a local configuration file for two devices using SNMP V2, with minimize_snmp set to true for the second device:

{
  "devices": [
    {
      "device_id": 2466,
      "snmp_comm": "device 2466 community string",
      "snmp_ip": "polling.ip.of.first_device"
    },
    {
      "device_id": 2681,
      "snmp_comm": "device 2681 community string",
      "snmp_ip": "polling.ip.of.second_device",
      "minimize_snmp": true
    }
  ]
}

When using SNMP v3, the additional required settings are specified with an snmp_v3 object. The object contains the following parameters:

An example snmp_v3 object is shown in the following sample code:

{
  "devices": [
    {
      "device_id": 3030,
      "snmp_comm": "device 3030 community string",
      "snmp_ip": "polling.ip.of.third_device",
      "minimize_snmp": false,
      "snmp_v3": "{\"UserName\":\"SNMPv3-Admin\",\"AuthenticationProtocol\":\"SHA\",\"AuthenticationPassphrase\":\"big_secret\",\"PrivacyProtocol\":\"AES\",\"PrivacyPassphrase\":\"unguessable\"}"
    }
  ]
}

The following tips may be useful in debugging issues related to the use of kproxy:

• Our article on Router Configuration will guide you through the general setup of routers to work with Kentik.
• The IPs allowed in the Agent tile on the Admin » Access Control page (see Access Control) must include the public IP of the server running kproxy. If the server is behind a NAT gateway you can get its public IP by running wget -qO- ifconfig.co on the server.
• If the kproxy command line argument -metrics was set to stderr then you will receive a checkpoint every minute that indicates how much flow you are receiving from the router. If that count is not increasing then there is an issue between your router and kproxy, either router configuration or kproxy config of communication between them.
• It may take 2-3 minutes for the agent to download flow templates and begin to process flow. You can expect to receive errors ("[ipfix_parse_msg] no template for 256, skip data set") for the first few minutes, after which the errors should stop.
• Errors will be logged in stdout by default, but if the -syslog flag was used in the kproxy command line then instead they are logged in syslog (see kproxy Command Line for details).
• A kproxy debug and health check port is opened by default on the loopback address (127.0.0.1), one port higher than the configured flow ingest port (e.g. for the default flow port of udp/9995, heathcheck is tcp/9996). The health check information fields are detailed in Health Check Fields.

The heath check output includes information about two classes of devices:

• Connected Devices: Devices that are already registered in Kentik and currently sending flow.
• Unregistered Devices: Devices that are not yet registered in Kentik but have sent flow in the last 30 minutes.

The health check output states the number of Connected Devices and returns the following fields for each:

• In1: Difference in incoming (from device) packet count over 1 minute.
• Out1: Difference in outgoing (to Kentik) packet count over 1 minute.
• In15: Difference in incoming (from device) packet count over 15 minutes.
• Out15: Difference in outgoing (to Kentik) packet count over 15 minutes.
• Last seen: Last time device sent a packet.
• Sources: IP addresses sending device packets.
• Channel highwater: Maximum lengths of internal message queues.

The health check output also states the number of Unregistered Devices and returns the following fields for each:

• Source IP address: Port sending flow packets.
• Timestamp: Time of first received flow packet.

The ingest of syslog data into KDE via kproxy is covered in the following topics:

• About Syslog Parsing
• Syslog Parsing Configuration
• Mapping Parsed Syslog
• Syslog Data Dimensions
• Using Syslog Data

In addition to its role as a collector and encryptor of flow data from routers and other network infrastructure, kproxy can also act as a parser for syslogs, enabling log data to be collected and stored in KDE alongside data from other Kentik-supported sources. This capability enables us to support monitoring and analytics in Kentik on a variety of information that may not be available directly from flow records. It works by matching patterns in the syslog text and assigning the resulting values to fields in a flow record that is ingested into KDE. Once the syslog data is stored in KDE, queries can filter and group-by on the dimensions corresponding to the fields populated from syslogs.

Because there is no universal standard for the structure of syslogs, Kentik enables the syslog parsing capability via a JSON configuration file that you use to tell us what patterns you want to match. If the path to such a file is specified with the syslog_config argument in the command line of a given kproxy instance (see kproxy Proxy Agent Arguments) then every syslog file collected by that instance will be evaluated for matches with the patterns defined in the config file.

Within the configuration file:

• The parsing instructions are structured as JSON.
• The patterns to match are defined using Grok. Learn more about Grok at:
  - https://logz.io/blog/logstash-grok/
  - https://www.elastic.co/guide/en/logstash/current/plugins-filters-grok.html
• A set of available default patterns is defined in the files found in the GitHub repo vjeantet/grok/patterns (https://github.com/vjeantet/grok/tree/master/patterns).
• A different set of patterns may be specified for each of the devices that send syslogs to a given instance of kproxy.

As shown in the commented configuration file example below:

• The patterns object allows you to specify custom (non-default) patterns that may be used to evaluate incoming syslogs from any device.
• The devices object identifies a device by its Kentik Device ID (generated when the device is onboarded; see Device List Columns) and determines how patterns are applied to that device.
• The types array in the devices object lists the patterns (default and custom) to apply for that device.

{
  "patterns": { /* A set of custom patterns to match (not needed when using only default patterns). */
    "IRCUSER": "\\A@(\\w+)",
    "IRCBODY": ".*",
    "IRCMSG": "%{IRCUSER:user}.* : %{IRCBODY:message}"
  },
  "devices": { /* A collection of objects that each represent the parsing settings for one device. */
    "1001": { /* An object, identified by Kentik device ID, containing the parsing settings for an individual device. */
      "skip_default": false, /* Determines whether matching of default patterns should (true) or should not (false) be skipped during parsing. */
      "types": [ /* An array of patterns to match for this device. If a pattern is not a default, it must be listed in the patterns object above. */
        "%{IRCMSG}", /* A pattern defined in the patterns section above. */
        "%{SHOREWALL}" /* A default pattern. */
      ]
    }
  }
}

Note: To use the above example as a configuration template, remove all comments.

As noted earlier, the purpose of parsing syslogs is to extract useful information that can be stored in the flow records of KDE. When a string in a parsed line of syslog is matched via the Grok configuration described above, the way that data is incorporated into a flow record depends on the factors described in the following topics.

The matched value will be entered into a standard KDE column (see Main Table Columns) in the following situations:

• The match is on a custom pattern that includes a label that matches a KDE column name. For example, if a pattern is defined as "(%{MY_PATTERN:patternlabel})" then Kentik would attempt to match the label "patternlabel" to a KDE column name. If a column by that name is not found then an attempt will be made to match the pattern as described in UDR Column Mapping.
• The match is on one of the default patterns for which there is both a label and a Kentik-defined mapping from the label to a KDE column. These currently include the following (more coming soon):
  - label "bytes" to column InBytes;
  - label "packets" to column InPkts;
  - label "clientip" to column DstIP;
  - label "localip" to column SrcIP.

If neither of the Standard Column Mapping situations apply, then the matching, if any, on the contents of a syslog line will result in data being entered into a UDR (custom) column (see Universal Data Records) rather than a standard KDE column:

• kproxy will only attempt to match patterns that are labeled as described in Standard Column Mapping.
• The labels of the patterns for which there are matches in a given syslog line will be ordered (ascending alphabetical).
• In the flow record for each syslog line, both the label and the matching data for each matched pattern will be assigned in order to one of the 11 available UDR fields, str00 through str10. For example if the first (alphabetically) pattern matched in a given syslog line is specified in the config file as "(%{MY_PATTERN:patternlabel})" then the UDR field str00 would be populated with "patternlabel = " followed by the text that matched the pattern MY_PATTERN.

Both Standard Column Mapping and UDR Column Mapping assume that kproxy has been configured with a -syslog_congig argument providing a config file (see kproxy Proxy Agent Arguments). In the absence of this configuration, kproxy takes a different approach to ingesting each syslog line into a KDE flow record:

• Characters 1-64 of the log line are assigned to field srt00.
• Characters 65-128 of the log line are assigned to field srt01.
• And so on (as needed) through field str10.

Note: A maximum of 704 characters will be ingested into the flow record via patternless mapping.

Once data from a syslog line has been ingested into a KDE flow record it is available for group-by and filtering in Kentik queries. The method used to access the data depends on how it was assigned to the fields of the flow record:

• Standard dimensions: If the matched data was put into a standard KDE column (see Standard Column Mapping) then it can be accessed via the dimension corresponding to that column (see Dimensions Reference).
• UDR dimensions: If the matched data was put into a UDR KDE column (str00 through str10) via either UDR Column Mapping or Patternless Mapping, then it can be accessed via the UDR dimension corresponding to that column. These dimensions are named Field00 through Field10.

Note: When filtering for data ingested via UDR Column Mapping you can use the label name that was inserted at the start of the field during ingest.

The setup workflow for using syslog data in Kentik is as follows:

1. Register with Kentik the devices that will be sending syslogs to kproxy (see kproxy Setup). The Type drop-down on the General tab of the Add Device dialog should be set to "Generic Syslog."
2. Create a configuration file that defines the patterns that you want to match and conforms to the structure described in Syslog Parsing Configuration:
   - If you are trying to ingest data that corresponds to an existing KDE column, see Standard Column Mapping.
   - If you are trying to ingest data for which there is no existing KDE column, see UDR Column Mapping.
3. Assign the configuration file to a given instance of kproxy using the -syslog_config command line argument (see kproxy Proxy Agent Arguments).
4. Once the devices begin sending syslogs to kproxy, the patterns defined in the configuration file will be used to match the contents of each syslog line and to map matching data to the fields of flow records.
5. When the flow records are ingested into KDE the syslog-derived data can be accessed via group-by and filtering on Syslog Data Dimensions.