Working with NetDevices

NetDevices is the core of Trigger’s device interaction. Anything that communicates with devices relies on the metadata stored within NetDevice objects.

Your Source Data

Before you can work with device metadata, you must tell Trigger how and from where to read it. You may either modify the values for these options within or you may specify the values as environment variables of the same name as the configuration options.

Please see configuration for more information on how to do this. There are two configuration options that facilitate this:

A URL or file path from which the metadata may be obtained. This defaults to /etc/trigger/netdevices.json, but can be any URL with variables.
(Advanced) A tuple of data loader classes, specified as strings. This is an advanced setting that you may use to create custom loaders if any of the default loaders do not meet your needs. The primary default loader is the JSONLoader.

A Brief Overview

When you instantiate NetDevices the location specified NETDEVICES_SOURCE is passed onto the NETDEVICES_LOADERS to try to parse and return device metadata.

Using the loaders, you don’t have to tell Trigger what the format of your metadata source is. It tries to determine it automatically based on whether one of the pre-defined loaders successfully returns data without throwing an error.

Anatomy of a Device

Trigger’s NetDevice objects represent everything Trigger needs to know about each device under its care. These objects are pretty complicated, but all you really need to know right now are the bare minimum set of fields that Trigger needs to know about your devices. These fields are used to control the behaviors and select the correct driver for each platform.

Field Values

Field values are expected to be strings. They are normalized prior to evaluation, so case-sensitivity can be left up to you based on how you choose to manage your data in your environment.

You may also specify custom fields not used or required by Trigger that you may use for your own purposes, such as writing custom utilities.

Minium Required Fields

These are the bare minimum required fields for basic operations of Trigger on most device platforms. Due to distinct differences across device platforms by vendor, hardware, and operating system versions, not all device platforms can be supported with the minimum felds.


Support for certain device platforms, such as the Cisco Nexus, require specifying other fields as detailed in the next section.

The device hostname or IP address. Trigger does not validate this value for you, so it must be able to be to be reconciled by the host system on which Trigger is running. You may also specify a port here by separating the hostname and port by a colon (e.g. hostname:2222) . We’ll cover that in more detail later.
The representative name of the hardware vendor. This is also used to dynamically populate the vendor attribute on the device object. For Trigger’s list of supported vendors, please see SUPPORTED_VENDORS.
(Recommended) The administrative status of the device. One of PRODUCTION or NON-PRODUCTION. By default, Trigger will only work with devices that are in PRODUCTION status. This is covered in more detail under Instantiating NetDevices. If you do NOT specifiy adminStatus, Trigger will fallback to the default value specified in DEFAULT_ADMIN_STATUS.
(Recommended) The type of device (e.g. router, switch, etc.). For the list of supported device types, please see SUPPORTED_TYPES. If you do NOT specify deviceType, Trigger will fallback to the type specified in FALLBACK_TYPE.

For example the minimum required fields for a device using JSON might be:

    "nodeName": "n9k1",
    "deviceType": "SWITCH",
    "manufacturer": "CISCO",
    "model": "NEXUS",
    "adminStatus": "PRODUCTION"

Common Fields

Some vendors, such as Cisco for example, have a wide array of hardware types with many different operating systems. For this reason, not all devices can be supported with just vendor and device type. To identify these platforms, we also make use of the make and model fields.

The following fields are required for full support of all officially supported vendor platforms.

The device platform such as Cisco Nexus or just Nexus.
The specific device model as it appears on the network device, such as N9K or NEXUS 9000.

Other Fields

There are a ton of other default fields that Trigger utilizes for other purposes.

Quick Start

To get started quickly managing real devices, try this:

  1. Create a CSV file.
  2. Tell Trigger where to find your file by setting the locatoin of the file in NETDEVICES_SOURCE in your

Importing from RANCID

Do you have RANCID? Try using that instead! To learn more please visit the section on working with the RANCID format.

Supported Formats

Trigger currently comes with loaders that support the following formats:

Except when using CSV or RANCID as a data source, the contents of your source data should be a dump of relevant metadata fields from your CMDB.

If you don’t have a CMDB, then you’re going to have to populate this file manually.


New in version 1.3.

This method is the most lightweight, but also the most limited. But it’s a great starting point!

The bare minimum config for CSV is a file populated comma-separated values, each on their own line with hostname,vendor. For example:,juniper,cisco

The most fields you may populate are the same as with the RANCID support. Please see the explanation of the fields populated by the RANCID format. A “fully-populated” CSV file would look more like this:,juniper,up,router,juniper,up,router,netscreen,up,firewall,foundry,up,switch


XML is the slowest method supported by Trigger, but it is currently the default for legacy reasons. At some point we will switch JSON to the default.

Here is a sample what the netdevices.xml file bundled with the Trigger source code looks like:

<?xml version="1.0" encoding="UTF-8"?>
<!-- Dummy version of netdevices.xml, with just one real entry modeled from the real file -->
    <device nodeName="">
        <budgetName>Data Center</budgetName>
        <lastUpdate>2010-07-19 19:56:32.0</lastUpdate>
        <make>M40 INTERNET BACKBONE ROUTER</make>
        <onCallName>Data Center</onCallName>
        <owningTeam>Data Center</owningTeam>
        <owner>12345678 - Network Engineering</owner>
        <projectName>Test Lab</projectName>

Please see configs/netdevices.xml within the Trigger source distribution for a full example.


JSON is the fastest method supported by Trigger. This is especially the case if you utilize the optional C extension of simplejson. The file can be minified and does not need to be indented.

Here is a sample of what the netdevices.json file bundled with the Trigger source code looks like (pretty-printed for readabilty):

        "adminStatus": "PRODUCTION",
        "enablePW": "boguspassword",
        "OOBTerminalServerTCPPort": "5005",
        "assetID": "0000012345",
        "OOBTerminalServerNodeName": "ts1",
        "onCallEmail": "",
        "onCallID": "17",
        "OOBTerminalServerFQDN": "",
        "owner": "12345678 - Network Engineering",
        "OOBTerminalServerPort": "5",
        "onCallName": "Data Center",
        "nodeName": "",
        "make": "M40 INTERNET BACKBONE ROUTER",
        "budgetCode": "1234578",
        "budgetName": "Data Center",
        "operationStatus": "MONITORED",
        "deviceType": "ROUTER",
        "lastUpdate": "2010-07-19 19:56:32.0",
        "authMethod": "tacacs",
        "projectName": "Test Lab",
        "barcode": "0101010101",
        "site": "LAB",
        "loginPW": null,
        "lifecycleStatus": "INSTALLED",
        "manufacturer": "JUNIPER",
        "layer3": "1",
        "layer2": "1",
        "room": "CR10",
        "layer4": "1",
        "serialNumber": "987654321",
        "owningTeam": "Data Center",
        "coordinate": "16ZZ",
        "model": "M40-B-AC",
        "OOBTerminalServerConnector": "C"

To use JSON, create your NETDEVICES_SOURCE file full of objects that look like the one above.

Please see configs/netdevices.json within the Trigger source distribution for a full example.


If you’ve already got a RANCID instance in your environment, this is the easiest method to get running. At this time, however, the metadata available from RANCID is very limited and populates only the following fields for each Netdevice object:

  • nodeName
  • manufacturer
  • deviceType
  • adminStatus

The support for RANCID comes in two forms: single or multiple instance.

Single instance is the default and expects to find the router.db file and the configs directory in the root directory you specify.

Multiple instance will instead walk the root directory and expect to find router.db and configs in each subdirectory it finds. Multiple instance can be toggled by seting the value of RANCID_RECURSE_SUBDIRS to True to your

To use RANCID as a data source, set the value of NETDEVICES_SOURCE in to the absolute path of location of of the root directory where your RANCID data is stored.


Make sure that the value of RANCID_RECURSE_SUBDIRS matches the RANCID method you are using. This setting defaults to False, so if you only have a single RANCID instance, there is no need to add it to your

Lastly, to illustrate what a NetDevice object that has been populated by RANCID looks like, here is the output of .dump():

Owning Org.:       None
Owning Team:       None
OnCall Team:       None

Vendor:            Juniper (juniper)
Make:              None
Model:             None
Type:              ROUTER
Location:          None None None

Project:           None
Serial:            None
Asset Tag:         None
Budget Code:       None (None)

Admin Status:      PRODUCTION
Lifecycle Status:  None
Operation Status:  None
Last Updated:      None

Compare that to what a device dump looks like when fully populated from CMDB metadata in What’s in a NetDevice?. It’s important to keep this in mind, because if you want to do device associations using any of the unpopulated fields, you’re gonna have a bad time. This is subject to change as RANCID support evolves, but this is the way it is for now.


SQLite is somewhere between JSON and XML as far as performance, but also comes with the added benefit that support is built into Python, and you get a real database file you can leverage in other ways outside of Trigger.

-- Table structure for table `netdevices`
-- This is for 'netdevices.sql' SQLite support within
-- trigger.netdevices.NetDevices for storing and tracking network device
-- metadata.
-- This is based on the current set of existing attributes in use and is by no
-- means exclusive. Feel free to add your own fields to suit your environment.

CREATE TABLE netdevices (
    OOBTerminalServerConnector VARCHAR(1024),
    OOBTerminalServerFQDN VARCHAR(1024),
    OOBTerminalServerNodeName VARCHAR(1024),
    OOBTerminalServerPort VARCHAR(1024),
    OOBTerminalServerTCPPort VARCHAR(1024),
    acls VARCHAR(1024),
    adminStatus VARCHAR(1024),
    assetID VARCHAR(1024),
    authMethod VARCHAR(1024),
    barcode VARCHAR(1024),
    budgetCode VARCHAR(1024),
    budgetName VARCHAR(1024),
    bulk_acls VARCHAR(1024),
    connectProtocol VARCHAR(1024),
    coordinate VARCHAR(1024),
    deviceType VARCHAR(1024),
    enablePW VARCHAR(1024),
    explicit_acls VARCHAR(1024),
    gslb_master VARCHAR(1024),
    implicit_acls VARCHAR(1024),
    lastUpdate VARCHAR(1024),
    layer2 VARCHAR(1024),
    layer3 VARCHAR(1024),
    layer4 VARCHAR(1024),
    lifecycleStatus VARCHAR(1024),
    loginPW VARCHAR(1024),
    make VARCHAR(1024),
    manufacturer VARCHAR(1024),
    model VARCHAR(1024),
    nodeName VARCHAR(1024),
    onCallEmail VARCHAR(1024),
    onCallID VARCHAR(1024),
    onCallName VARCHAR(1024),
    operationStatus VARCHAR(1024),
    owner VARCHAR(1024),
    owningTeam VARCHAR(1024),
    projectID VARCHAR(1024),
    projectName VARCHAR(1024),
    room VARCHAR(1024),
    serialNumber VARCHAR(1024),
    site VARCHAR(1024)

To use SQLite, create a database using the schema provided within Trigger source distribution at configs/netdevices.sql. You will need to populate the database full of rows with the columns above and set NETDEVICES_SOURCE the absolute path of the database file.

Developing with NetDevices

First things first, you must instantiate NetDevices. It has three things it requires before you can properly do this:


If you do not want to load ACL associations you may skip them by passing with_acls=False to NetDevices and then you only need to satisfy the first requirement. A this time it is not possible to globally disable ACL support, so this will only work for the purpose of this walkthrough or when you manually instantiate NetDevices objects.

  1. The NETDEVICES_SOURCE file must be readable and must properly parse using one of the default loaders formats supported in NETDEVICES_LOADERS (see above);
  2. An instance of Redis (you may skip this by passing with_acls=False to the NetDevices constructor).
  3. The path to must be valid, and must properly parse (you may skip this if you just want to ignore the warnings for now).

How it works

The NetDevices object itself is an immutable, dictionary-like Singleton object. If you don’t know what a Singleton is, it means that there can only be one instance of this object in any program. The actual instance object itself an instance of the inner _actual class which is stored in the module object as NetDevices._Singleton. This is done as a performance boost because many Trigger components require a NetDevices instance, and if we had to keep creating new ones, we’d be waiting each time one had to parse NETDEVICES_SOURCE all over again.

Upon startup, each device object/element/row found within NETDEVICES_SOURCE is used to create a NetDevice object. This object pulls in ACL associations from AclsDB.

The Singleton Pattern

The NetDevices module object has a _Singleton attribute that defaults to None. Upon creating an instance, this is populated with the _actual instance:

>>> nd = NetDevices()
>>> nd._Singleton
<trigger.netdevices._actual object at 0x2ae3dcf48710>
>>> NetDevices._Singleton
<trigger.netdevices._actual object at 0x2ae3dcf48710>

This is how new instances are prevented. Whenever you create a new reference by instantiating NetDevices again, what you are really doing is creating a reference to NetDevices._Singleton:

>>> other_nd = NetDevices()
>>> other_nd._Singleton
<trigger.netdevices._actual object at 0x2ae3dcf48710>
>>> nd._Singleton is other_nd._Singleton

The only time this would be an issue is if you needed to change the actual contents of your object (such as when debugging or passing production_only=False). If you need to do this, set the value to None:

>>> NetDevices._Singleton = None

Then the next call to NetDevices() will start from scratch. Keep in mind because of this pattern it is not easy to have more than one instance (there are ways but we’re not going to list them here!). All existing instances will inherit the value of NetDevices._Singleton:

>>> third_nd = NetDevices(production_only=False)
>>> third_nd._Singleton
<trigger.netdevices._actual object at 0x2ae3dcf506d0>
>>> nd._Singleton
<trigger.netdevices._actual object at 0x2ae3dcf506d0>
>>> third_nd._Singleton is nd._Singleton

Instantiating NetDevices

Throughout the Trigger code, the convention when instantiating and referencing a NetDevices instance, is to assign it to the variable nd. All examples will use this, so keep that in mind:

>>> from trigger.netdevices import NetDevices
>>> nd = NetDevices()
>>> len(nd)

By default, this only includes any devices for which adminStatus (aka administrative status) is PRODUCTION. This means that the device is used in your production environment. If you would like you get all devices regardless of adminStatus, you must pass production_only=False to the constructor:

>>> from trigger.netdevices import NetDevices
>>> nd = NetDevices(production_only=False)
>>> len(nd)

The included sample metadata files contains one device that is marked as NON-PRODUCTION.

What’s in a NetDevice?

A NetDevice object has a number of attributes you can use creatively to correlate or identify them:

>>> dev = nd.find('test1-abc')
>>> dev

Printing it displays the hostname:

>>> print dev

You can dump the values:

>>> dev.dump()

        Owning Org.:       12345678 - Network Engineering
        Owning Team:       Data Center
        OnCall Team:       Data Center

        Vendor:            Juniper (JUNIPER)
        Make:              M40 INTERNET BACKBONE ROUTER
        Model:             M40-B-AC
        Type:              ROUTER
        Location:          LAB CR10 16ZZ

        Project:           Test Lab
        Serial:            987654321
        Asset Tag:         0000012345
        Budget Code:       1234578 (Data Center)

        Admin Status:      PRODUCTION
        Lifecycle Status:  INSTALLED
        Operation Status:  MONITORED
        Last Updated:      2010-07-19 19:56:32.0

You can reference them as attributes:

>>> dev.nodeName, dev.vendor, dev.deviceType
('', <Vendor: Juniper>, 'ROUTER')

There are some special methods to perform identity tests:

>>> dev.is_router(), dev.is_switch(), dev.is_firewall()
(True, False, False)

You can view the ACLs assigned to the device:


If you have passed with_acls=False, none of these attributes will be populated and will instead return an empty set()).

>>> dev.explicit_acls
>>> dev.implicit_acls
set(['juniper-router.policer', 'juniper-router-protect'])
>>> dev.acls
set(['juniper-router.policer', 'juniper-router-protect', 'abc123'])

Or get the next time it’s ok to make changes to this device (more on this later):

>>> dev.bounce.next_ok('green')
datetime.datetime(2011, 7, 13, 9, 0, tzinfo=<UTC>)
>>> print dev.bounce.status()

Searching for devices

Like a dictionary

Since the object is like a dictionary, you may reference devices as keys by their hostnames:

>>> nd
{'': <NetDevice:>,
 '': <NetDevice:>,
 '': <NetDevice:>,
 '': <NetDevice:>}
>>> nd['']

You may also perform any other operations to iterate devices as you would with a dictionary (.keys(), .itervalues(), etc.).

Special methods

There are a number of ways you can search for devices. In all cases, you are returned a list.

The simplest usage is just to list all devices:

>>> nd.all()
[<NetDevice:>, <NetDevice:>,
 <NetDevice:>, <NetDevice:>]

Using all() is going to be very rare, as you’re more likely to work with a subset of your devices.

Find a device by its shortname (minus the domain):

>>> nd.find('test1-abc')

List devices by type (switches, routers, or firewalls):

>>> nd.list_routers()
[<NetDevice:>, <NetDevice:>]
>>> nd.list_switches()
>>> nd.list_firewalls()

Perform a case-sensitive search on any field (it defaults to nodeName):

[<NetDevice:>, <NetDevice:>]
>>>'NON-PRODUCTION', 'adminStatus')

Or you could just roll your own list comprehension to do the same thing:

>>> [d for d in nd.all() if d.adminStatus == 'NON-PRODUCTION']

Perform a case-INsenstive search on any number of fields as keyword arguments:

>>> nd.match(oncallname='data center', adminstatus='non')
>>> nd.match(vendor='netscreen')

Helper function

Another nifty tool within the module is device_match, which returns a NetDevice object:

>>> from trigger.netdevices import device_match
>>> device_match('test')
2 possible matches found for 'test':
 [ 1]
 [ 2]
 [ 0] Exit

Enter a device number: 2

If there are multiple matches, it presents a prompt and lets you choose, otherwise it chooses for you:

>>> device_match('fw')
Matched ''.