For vendors and partners

Offboarding a Client From Domain-Separated ServiceNow (MSP Guide)

Offboarding a client from domain-separated ServiceNow: an MSP guide to filtering by sys_domain, untangling global data and moving one tenant out clean.

Ayush BothraFounding Member, MigrateX

Published 15 min read

Geometric art of one colored group of blocks lifted cleanly out of a shared grid, for a ServiceNow domain separation migration guide
On this page

To move one customer out of an instance that uses ServiceNow domain separation, you extract that tenant's records by their sys_domain value, then follow every reference to the shared global data they depend on: users, assignment groups, companies, configuration items (CIs) and the rules behind them. Done right, the new platform gets everything that belongs to the tenant and nothing that belongs to anyone else.

This guide is for implementation partners, MSPs and systems integrators: you are either winning a customer away from another provider's shared ServiceNow, or offboarding a customer from your own. If you are the customer stuck in a shared instance, it is for you too. It shows what to ask your provider for, and what a clean exit looks like.

What is ServiceNow domain separation, and why does it make migration hard?

ServiceNow domain separation lets one instance serve many customers by tagging records with a domain, stored in the sys_domain field, and showing each user only their own domain plus the domains above it. MSPs use it to run many tenants on one instance. It makes migration hard because tenants share a global layer of data and logic.

Three rules from ServiceNow's documentation explain most of the migration work:

  • Visibility flows down, not up. "A record in the parent domain can be seen in the child, but a parent cannot see a child record" (ServiceNow docs, Domain separation terms). Anyone can see records in the global domain.
  • Records take a domain when they are created. The domain comes from the user's company, a business rule, the module or form template used, the parent record, or the user record. "If no other domain conditions apply, a record automatically inherits the domain of the user who creates it" (ServiceNow docs, Domain assignment).
  • Logic is inherited and overridden. Business rules, notifications, client scripts and UI policies can be set per domain. A domain-specific copy is a process override, flagged in the Overrides [sys_overrides] field (ServiceNow docs, Domain separation terms).

So a tenant is never just "the records in one domain". It is those records, plus the global records they point at, plus the global logic that shapes them, minus whatever overrides the tenant had. A migration that moves only the first part arrives broken. One that moves too much leaks other customers' data.

What lives in the global domain and what lives in the tenant domain?

In a typical MSP instance, the tenant's tickets, requests, changes, problems and their task SLAs sit in the tenant domain. Shared reference data and logic sit in global or in the MSP's own domain. The exact split varies by instance, so treat this table as a starting map and confirm every row against the real data.

DataUsually lives inMigration riskWhat to do
Incidents, requests, problems, changesTenant domain (and its child domains)Low if filtered correctlyPull by sys_domain, including every child domain
Task SLAsTenant domain: ServiceNow creates them "only in the domain of its attached task record" (ServiceNow docs, Domain separation and Service Level Management)LowPull with their parent tasks
Journal entries, audit history, attachmentsPlatform tables that hang off the parent recordHigh: easy to miss or over-pullSelect by parent record sys_id, never by a domain filter alone
End users (callers, requesters)Tenant domain. The docs say "user record data never belongs in the global domain", but admins report finding users thereMediumPull tenant users; review any global user a tenant ticket references
Agents and resolver groupsOften the MSP's own domain or global, because one service desk works many tenantsHigh: tenant tickets point at people who are not the tenant'sMap to target groups, or keep the name as history
Companies, locations, departmentsTenant company in its domain; vendors and the MSP itself often globalMediumCopy the tenant's own; map shared ones
Configuration itemsTenant CIs in its domain; shared hosting and network CIs often global or MSP domainHigh: relationships cross domainsPull tenant CIs and relationships; decide per shared CI
Knowledge articles, catalog itemsMix of global and tenantMediumMove tenant articles; list global ones the tenant actually used
SLA definitionsGlobal or tenant; ancestor definitions "can be overridden in a sub-domain" (ServiceNow docs, Domain separation and Service Level Management)High: the effective SLA is the override, not the global oneDocument the effective definition per tenant
Business rules, notifications, client scripts, UI policiesGlobal, with domain-specific overridesHighTreat as a rebuild spec, not data
Choice lists, categories, dictionaryGlobal platform tables; ServiceNow advises against domain-separating sys_ tablesMediumExport the values the tenant's records actually use

Can ServiceNow's Domain Migration Tool move a tenant?

Only to a dedicated ServiceNow instance. ServiceNow's Domain Migration Tool runs on a clone of the shared instance. It deletes every domain except the target tenant and any extra data domains you name, then switches domain separation off. It cannot send a tenant to another MSP's instance or to a different platform.

The facts, from ServiceNow's documentation for the Domain Migration Tool:

  • You need a clone and a plugin. "You must request a cloned instance and request activation of the Domain Migration Tool plugin before you can use it" (plugin com.glide.domain.migration_tool). A reply in a January 2025 ServiceNow Community thread describes it as a paid plugin that ServiceNow staff activate.
  • Both separation properties must be on. glide.sys.domain.partitioning and glide.sys.domain.delegated_administration must be true. In a March 2022 Community thread, a customer reported that "our MSP has blocked the property". A ServiceNow employee suggested cloning production to a sub-production instance and enabling it there. Either way, the provider has to cooperate.
  • Global data stays. "The tool doesn't delete data that is global, in the target domain, or in additional data domains." Anything other tenants left in global comes with the clone.
  • Configuration needs a review. You "must evaluate all the remaining configurations" afterwards, for example business rules that set the domain field.

If your client wants their own ServiceNow instance and the current provider will cooperate on the clone, the tool is a reasonable route. The questions partners actually bring us are different: move this tenant into our domain-separated instance, or into Freshservice or Jira Service Management. A 2018 ServiceNow Community thread asking how to move one domain out of an MSP instance ended without a method. The asker concluded a fresh build might cost less.

Which extraction method fits: Domain Migration Tool, IDR, XML or API?

Pick by target first. The Domain Migration Tool only produces a dedicated ServiceNow instance. Instance Data Replication (IDR) and XML export only reach another ServiceNow instance. API extraction is the only route that reaches any target, including another provider's domain-separated instance, with the tenant's full history.

Domain Migration ToolInstance Data Replication (IDR)XML export and importAPI extraction (how MigrateX works)
TargetA new dedicated ServiceNow instance, built from a cloneAnother ServiceNow instanceAnother ServiceNow instanceAny target: another MSP's ServiceNow, a dedicated instance, Freshservice, Jira Service Management and others
ScopeKeeps global, the target domain and named data domains; deletes the restTables and filters you configure. ServiceNow's docs say IDR "does not explicitly support domain separation"Whatever a list filter returns, one table at a timeThe tenant's domains, plus only the global records its data references
HistoryWhatever is in the cloneReplicates record updates from producer to consumer; built for ongoing syncList XML export includes work notes and comments; URL export leaves out "attachments or journal fields". Audit history needs its own exportJournal entries, audit history and attachments, selected by parent record
Needs the current provider toRequest the clone and enable two propertiesSet up a producer instanceGrant list export rightsGrant a read-only API user
CostPaid plugin, per a ServiceNow Community threadPaid plugin, per a ServiceNow Community thread from October 2025, where replies also disagree on whether it suits a one-time moveNo licence; heavy manual effort, plus export limitsFixed quote after a free 100-record sample
Best forCustomer becomes its own ServiceNow ownerOngoing sync between instancesSmall, one-off data setsTenant moves to another provider or platform

Our guide to exporting every record from ServiceNow covers export limits, Table API paging and attachment tables in more depth.

How do you extract one tenant from a shared ServiceNow instance?

Use a read-only API user, map the tenant's domain tree, and pull its records by sys_domain. Then follow every reference outward, collect child data by parent record, and capture the logic as a rebuild spec. Run a leak check before anything loads. These ten steps are the order we follow.

  1. Agree access with the current provider. Ask for a read-only integration user. Where the provider allows it, place that user in the tenant's domain, not global: a parent cannot see a child, so the user cannot see sibling tenants even if a filter is wrong. Get this agreed in writing before notice periods start.
  2. Map the domain tree. Read the domain table and list the tenant's domain plus every child domain (sites, business units, subsidiaries). A filter on the top domain alone misses tickets logged in a child.
  3. Inventory the tables. For each table with a sys_domain field, count the tenant's records by type, status and year. These counts become the baseline for reconciliation.
  4. Pull the tenant's records. Query each table with a sys_domain filter over the full domain list, paging with the Table API's sysparm_limit and sysparm_offset. Ask for display values and raw sys_ids together, so references can be resolved later.
  5. Walk the references outward. For every reference field (caller, assigned to, assignment group, company, location, configuration item), collect the sys_ids that resolve outside the tenant's domains. This is the global dependency list, and every item on it needs a decision: copy, map to a target equivalent, or keep as a name only.
  6. Pull child data by parent. Collect work notes and comments, audit history, attachments, task SLAs, approvals, task-to-CI links and CI relationships by the parent record's sys_id. These tables do not carry a reliable tenant filter of their own.
  7. Capture the logic as a spec. List the business rules, notifications, SLA definitions, client scripts and catalog items that apply to the tenant, and record the overridden version wherever a sys_overrides value exists. Logic is rebuilt on the target, not copied as data.
  8. Run a leak check. Before loading, scan the extract for any sys_domain outside the tenant's list, and for any company, group or user sys_id the dependency decisions did not approve. Anything found is removed or justified in writing.
  9. Load under the right identity. Into a domain-separated target, load with a user in the target tenant's domain, or set the domain explicitly on every record. Never load as the global admin. Into Freshservice or Jira Service Management, map the tenant's companies, groups and categories to the target's model first.
  10. Reconcile, sync and cut over. Compare counts against step 3, diff fields on a sample, check files and links. Then run a final sync of records that changed since the bulk load, and cut over with a rollback plan ready. Our data migration validation guide covers the sign-off report.

What should partners scope before quoting a tenant extraction?

Scope the access, the domain tree, the shared data and the logic before you quote. Most overruns on tenant extractions come from one of four places: the provider will not grant access, child domains nobody listed, a global dependency list far longer than expected, or SLAs that depend on overrides nobody documented.

Access and contract

  • Who owns the instance and the licences: the provider, a reseller, or the customer?
  • Will the provider grant a read-only API user, and in which domain?
  • What does the contract say about data return, format and the date access ends?
  • Is there a notice period or exit clause that sets the cutover date?

Domain structure

  • How many domains belong to the tenant, including child domains?
  • Did the tenant ever change name or merge, leaving records in an old domain?
  • Are there custom tables, and do they carry sys_domain?

Data and history

  • Record counts per table, by status and year.
  • How many years of history are in scope: work notes, comments, audit history?
  • Attachment count and total size.
  • Archived records in ar_ tables.

Shared dependencies

  • How many tenant tickets reference agents or groups outside the tenant's domains?
  • Which CIs are shared (hosting, network, MSP tooling), and do the tenant's relationships point at them?
  • Which knowledge articles and catalog items does the tenant use from global?

Logic

  • Which SLA definitions apply to the tenant, and are any overridden?
  • Which business rules, notifications and workflows have tenant overrides?
  • Which integrations (email, monitoring, single sign-on) are configured per tenant?

Target

  • Is the target a domain-separated ServiceNow instance, a dedicated instance, or another platform?
  • How will the target model the MSP agents who worked the old tickets?

Our post on protecting project margins as an SI covers how to price the unknowns that remain.

What breaks when a tenant is moved out?

What breaks is almost always a link across the domain boundary: a ticket pointing at a group that did not move, a CI relationship to shared infrastructure, an SLA whose real definition was an override. These are the failures we check for on every domain-separated move.

  • Assignment history points at nobody. Tickets worked by the old provider's shared service desk reference agents and groups that do not exist on the target.
  • Child-domain tickets go missing. The filter covered the top domain, not the sites or business units below it.
  • SLA results change. The target rebuilt the global SLA definition, not the tenant's override, so historic breaches no longer match the reports the customer remembers.
  • CI relationships are cut. The tenant's servers depended on shared network or hosting CIs that stayed behind.
  • Records land in the wrong domain. A load run as global admin puts the tenant's data in global, visible to every tenant on the new instance.
  • Other tenants' data leaks in. A global admin extract with a missed filter, or a dependency walk that copied whole shared tables.
  • Work notes and attachments are missing. They were filtered by domain instead of by parent record, or left out of a URL-based XML export.
  • Notifications fire on old tickets. The target's rules ran during the load and emailed requesters about years-old incidents.
  • Choice values go blank. Categories or states that existed only as global choices were never mapped.
  • The provider's access ends first. The extract started after the notice period was already running.

Our test migration checklist shows how to catch these in a test run before the production cutover.

How does white-label delivery work for partners?

You stay in front of the client; we run the move under your name. You own the relationship, the scoping call and the sign-off. MigrateX works behind your team: we connect to the source instance, extract and load the tenant, and hand you the reconciliation report to deliver as your own.

In practice:

  • Your brand on everything the client sees. Plans, status updates and the reconciliation report go out under your name. We join client calls as part of your team, or not at all. Your choice.
  • Scoping before you quote. We run the free 100-record sample from the tenant and give you a fixed quote within 24 hours, so your proposal carries a firm migration number.
  • Any target. Into your own domain-separated ServiceNow, a dedicated ServiceNow instance, Freshservice, Jira Service Management or another of our 191 live connectors. If a platform is missing, we build the connector in 2 to 3 days.
  • Clean handover. You get the counts, the exceptions log and the global dependency decisions, so your team can support the client from day one.

One example of a partner-led ServiceNow move: Metro Systems, an IT services partner in Thailand, delivered a ServiceNow to Freshservice migration for an energy group through us. The full run moved 19,970 records in 136 minutes. Natchaporn (Neena) Nukulpong, Project Manager at Metro Systems: "Our customer was moving from ServiceNow to Freshservice, and the go-live date was brought forward. Within 12 hours they had mapped the fields with us and shown our customer a demo to check. The full migration took just over two hours. Our customer went live on time." (Metro Systems story)

For scale, Covia, an industrial minerals company in the US and Mexico, moved from ServiceNow to Freshservice with us: 1.03M+ records with full history, 369K+ of them in one weekend, covering 9 years of English and Spanish records (Covia story).

How does MigrateX handle a domain-separated ServiceNow extraction?

MigrateX is an AI migration platform, and ITSM experts review every step. For a shared instance, that means we extract by domain, walk the global dependencies, pull history by parent record, and leak-check the extract before anything loads. Then we reconcile every record and hand over a report.

  • Tenant-scoped extraction. Records pulled by the tenant's full domain list, with the global dependency list reviewed with you, item by item.
  • Full history. Work notes, comments, audit history and attachments come with their parent records, with original authors and timestamps.
  • ServiceNow to ServiceNow, or anywhere else. Instance-to-instance moves into another domain-separated or dedicated ServiceNow, or into any of our 191 live connectors. See the ServiceNow platform page and the ServiceNow to Freshservice and ServiceNow to Jira Service Management routes.
  • 100% data fidelity, proven. Counts, field diffs and file checks reconciled source to target, with every exception closed before sign-off.
  • Final sync and rollback plan. Changes made after the bulk load are picked up at cutover, and every production cutover has a rollback plan ready.
  • Security your client can review. SOC 2 Type II, GDPR and HIPAA compliant, with a DPA available (see security). Trusted by 1,000+ companies globally.

What to have ready for the call

You do not need all of this; whatever you have makes the quote faster.

  • Source instance owner (provider, reseller or customer) and the contract end or notice date
  • Tenant record counts by table, or a rough ticket count and years of history
  • Number of tenant domains, including child domains, if known
  • Target platform and whether it is domain-separated
  • Whether the provider will grant a read-only API user
  • Any SLA, CMDB or knowledge content the customer cannot lose

Facts checked November 11, 2026 from ServiceNow docs: Domain Migration Tool, ServiceNow docs: Domain separation terms, ServiceNow docs: Domain assignment, ServiceNow docs: Domain separation and Service Level Management, ServiceNow docs: Domain separation and IDR, ServiceNow docs: Export directly from a URL, ServiceNow Table API, ServiceNow Community: migrating from a domain-separated MSP instance (2018), ServiceNow Community: MSP blocked a required property (2022), ServiceNow Community: Domain Migration Tool cost (2025), ServiceNow Community: IDR for one-time migration (2025) and Precision Bridge migration guide. No vendor paid for this article. How we research

ServiceNow domain separation migration FAQ

What is domain separation in ServiceNow?

Domain separation is a ServiceNow feature that partitions one instance into domains, so each customer or business unit sees only its own data and its own versions of rules and processes. Each record's domain is stored in the sys_domain field. Records in a parent domain are visible to its children, and records in global are visible to everyone.

Can I move one tenant out of an MSP's ServiceNow instance?

Yes. Pull the tenant's records by sys_domain across its full domain tree, follow references to the global users, groups, companies and CIs they depend on, and collect history and attachments by parent record. The current provider has to grant access, so agree a read-only API user before the notice period starts.

Does the Domain Migration Tool work for moving to another MSP or to Freshservice?

No. Per ServiceNow's documentation, the Domain Migration Tool converts a clone of the shared instance into a new dedicated ServiceNow instance. It keeps global data, the target domain and any named data domains, and deletes the rest. Moving a tenant to another provider's instance or another platform needs API extraction.

Why do migrated records end up in the global domain?

Because a new record takes its domain from the user who creates it when nothing else sets it, and admin accounts usually sit in global. Precision Bridge's guide warns never to migrate as the global admin user for this reason. Load with a user in the target tenant's domain, or set the domain on each record explicitly.

Will other customers' data come across by mistake?

It can, if the extract runs as a global admin with a missed filter, or if whole shared tables are copied. Use a read-only user in the tenant's domain where possible, copy only the global records the tenant's data references, and scan the extract for any domain outside the tenant's list before loading.

Do work notes, audit history and attachments come with the tenant?

They can, but not through a domain filter alone. Journal entries, audit records and attachments sit in platform tables linked to the parent record, so they are collected by the parent's sys_id. A URL-based XML export leaves out attachments and journal fields, per ServiceNow's docs.

Can MigrateX work under our brand as the partner?

Yes. We deliver white-label: you stay in front of the client, and we run the extraction and load under your name. You get a free 100-record sample from the tenant and a fixed quote within 24 hours to put in your proposal.

Get the tenant out cleanly

Partners and MSPs: register the project or talk to our partner team. Send the tenant's record counts and the target platform, and we will scope the extraction with you, white-label. See how we work with partners.

Stuck in a shared instance? Ask for a free 100-record extract from your tenant, including work notes, audit history and attachments, and a fixed quote within 24 hours. Request your sample.

Related reading: leaving Kyndryl's ServiceNow with your data, merging two ServiceNow instances, Flow by ServiceNow and your ticket data, and our ServiceNow migration guide.

Changelog

  • 11 Nov 2026: First published.

Written by

Ayush Bothra

Founding Member, MigrateX

Ayush writes about ITSM platforms, pricing and what it takes to move a service desk to a new platform without losing its history.