CMDB / Integrations
Bringing Fleet device data into the ServiceNow CMDB
Fleet already knows which computers you run, what is installed on them and who uses them. Getting that into the ServiceNow CMDB is less about moving data than about a handful of decisions: which class each device belongs in, how ServiceNow recognizes a device it has seen before, which source wins when Fleet, SCCM, Intune or Jamf disagree, where software lands, and what happens when a device goes quiet.
Why Fleet data belongs in the CMDB
Fleet, also known as FleetDM, is a device management platform built on osquery. Its agent reports hardware, operating system, network and software details from macOS, Windows and Linux machines, and Fleet MDM extends that to iPhones, iPads and Android devices. For the endpoint team, Fleet is often the most current picture of the environment.
Auditors and asset reviews ask a different question: what does the system of record say? When the CMDB and Fleet hold two separate inventories, every review turns into a reconciliation exercise done by hand. Bringing Fleet into ServiceNow means the CMDB can answer with data that is as current as the endpoint tool, and that incident, change and asset processes use the same records.
The integration itself is the easy part. The decisions underneath it decide whether the CMDB ends up trusted or full of duplicates, and they are the same decisions whichever tool you use to move the data.
Decide what becomes a CI, and in which class
Start with a class map. Laptops and desktops usually belong in the computer class, servers in the server class for their operating system, and mobile devices in their own hardware classes. Every class carries its own identification, reconciliation and health rules, so a device in the wrong class is matched and measured by the wrong rules.
Then decide what stays operational detail. Fleet can report far more than a CMDB needs: running processes, every package on a Linux server, every browser extension. Load what an asset, incident or audit process will actually use. Running processes, for example, are useful for a short list of named agents or services and noise for everything else, so treat them as opt-in.
Mobile devices deserve their own decision. If phones and tablets are already tracked as assets from another source, adding them from Fleet MDM can create a second copy. Decide which source owns them before switching them on.
Identification: serial numbers, with their type
ServiceNow's Identification and Reconciliation Engine, IRE, decides whether an incoming record is a device it already holds. For hardware, the most reliable identifier is usually the serial number, and the out-of-box rules can match on a serial number stored in the serial number table together with its type. Recording the type matters: a system serial, a chassis serial and a board serial are different values, and matching one source's chassis serial against another's system serial finds nothing.
Fleet reports the hardware serial for each host. Writing it with its type lets the same identification rules match the device against records from SCCM, Intune, Jamf or Discovery, instead of creating a second CI for the same laptop. Names are a weak fallback: one source sends a short hostname, another a fully qualified one, and machines get renamed.
Virtual machines and cloned images are the usual exception, because serials can be blank or copied from a template. Check what your virtual hosts actually report before relying on serial number alone for those classes, and keep a secondary identifier, such as a network adapter's MAC address, in the rule.
Reconciliation: running alongside SCCM, Intune and Jamf
Most environments have more than one source that knows about the same device. Fleet, Microsoft Configuration Manager, Intune and Jamf may all report the same Mac or Windows laptop. Without rules, the last import to run wins, and the operating system version or memory on a CI changes with every load.
Give each integration its own discovery source value, so ServiceNow knows which source sent each update. Then use reconciliation rules to say which source is authoritative for which classes and attributes, data source precedence to rank them, and data refresh rules to let a lower-ranked source take over a value once the authoritative one has stopped reporting for long enough.
This only works if every write goes through IRE. An older import set whose transform map writes directly to the CI tables bypasses identification and reconciliation entirely, so its records can duplicate or overwrite what the rules protect. When you add Fleet, make sure its data takes the IRE path, and check the other sources do too.
- Which classes may Fleet create, and which may it only update?
- For which attributes is Fleet the best source in your environment, and for which is another tool better?
- How long should a device go unreported before another source's values take over?
Software: install records, publishers, and honest gaps
Where installed software lands depends on whether Software Asset Management is in use. With SAM, installations belong in software install records, where normalization and license reconciliation can use them. Without SAM, they have traditionally been held as software instances related to the device, but recent releases route more discovered software to the install table even without full SAM, so check which table your instance actually uses. Pick one path and keep every source on it, or the same install is counted twice.
Normalization depends on knowing the publisher. Fleet's software catalog carries publisher information for many titles, and passing it through gives SAM a better chance of matching each install to the right product.
Be honest about what the source does not provide. Fleet's API does not return a software install date, so the safest choice is to leave that field blank. Filling it with the date the integration first saw the install looks like data but is not, and it will mislead the first audit that relies on it. The last-used date comes from Fleet's last-opened data, which is reliable on macOS and indicative on Windows and Linux; treat it accordingly before using it to reclaim licenses.
Users, stale devices and retirement
Fleet knows who a device is assigned to. Match that person to an existing ServiceNow user and leave the field empty when there is no match. Creating users from device data fills the user table with duplicates and leavers, and identity belongs to your identity source, not your endpoint tool.
Devices stop reporting for ordinary reasons: a laptop in a drawer, a reimaged machine, a server switched off for maintenance. Deleting the CI when that happens destroys its history and its relationships to incidents, changes and assets. Mark it stale instead, and make retirement a deliberate step that follows your lifecycle policy, for example through CMDB Data Manager policies rather than an import that deletes what it no longer sees.
Make the mapping visible and the runs auditable
A mapping hidden in a script is one person's knowledge. IntegrationHub ETL lets the ServiceNow team see how each Fleet field maps to a class and attribute, adjust it, and preview what IRE will do with a record before it is loaded.
Run full imports on a schedule to catch drift, and change-only imports in between to keep data current without reloading everything. Keep a log of every run with its counts of records created, updated and skipped. When an auditor asks where a value came from and when, that log is the answer.
Use a read-only API user in Fleet. The integration only needs to read, and a CMDB feed should never be able to change device management.
How the Emetrix connector applies these decisions
The Emetrix CMDB Connector for Fleet is in early access. Version 1.0 is built to bring Fleet devices into the CMDB as configuration items with serial numbers recorded with their type, every write going through IRE, and the mapping built in IntegrationHub ETL so your team can see and adjust it. It ships recommended reconciliation rules and its own discovery source, so it can run alongside SCCM, Intune, Jamf and Fleet's own sync.
Installed software goes to SAM install records when SAM is in use and to software instances otherwise, with publishers from Fleet's catalog where available. The install date is left blank, users are matched and never created, devices that stop reporting are marked stale and never deleted, and running processes are off by default. It reads Fleet with a read-only API user and works with Fleet 4.70 and later, on Fleet Free or Premium.
Before connecting Fleet to ServiceNow, write down the class for each kind of device, the identifiers each source really sends, which source is authoritative for which attributes, where software installs belong, and what happens when a device stops reporting. Those answers decide whether the CMDB becomes evidence you can show an auditor or another inventory to reconcile.
References
- Identification and Reconciliation Engine (IRE) (opens in a new tab)ServiceNow documentation
- Reconciliation rules (opens in a new tab)ServiceNow documentation
- IntegrationHub ETL (opens in a new tab)ServiceNow documentation
- REST API (opens in a new tab)Fleet documentation
Related reading
- CMDB and CSDM at EmetrixCMDB trust first, then CSDM alignment, starting with a CMDB Trust Check.
- Duplicates that came back every quarterField note, a composite pattern rather than a client.
- Scoped app, product, or extension? Deciding before anyone writes codeArticle on deciding between a scoped application, a product and an extension.
Want this looked at on your platform?
CMDB Trust Check: Typically two to three weeks.
Ask about a CMDB Trust CheckWe reply within 24 hours on business days.