OBLYTECH Logo
OBLYTECH Logo
ServiceNow Upgrade Readiness Automation Opportunities

ServiceNow Upgrade Readiness Automation Opportunities

Upgrade automation means scripting the repeatable checks and steps around a ServiceNow upgrade. The page covers compatibility scans of customizations and plugins, data validation, backups, staging tests, multi-system coordination, cutover and rollback, and post-upgrade validation. Each check should produce a report that the development team can review before the upgrade window.

Automate Compatibility Checks to Identify Plugin and Customization Conflicts

Before upgrading ServiceNow, you must verify that your customizations and plugins are compatible with the new release. Many conflicts are predictable: deprecated APIs, removed tables, changed field types. Use compatibility scanning tools to analyze your customizations and identify potential issues before upgrade. Common issues include customizations that conflict with changes in the new release, which appear on the skipped records list. Automate this scanning in your release readiness workflow: scan your configuration, generate a report of potential issues, and route the report to the development team for review. This early identification allows time to fix issues before the upgrade window rather than discovering problems during the upgrade and having to roll back.

Automate Data Validation and Recovery Planning Before Upgrade Execution

An upgrade affects your entire ServiceNow instance and the data in it. Before upgrading, validate that your data is in a good state: records have required fields populated, no orphaned references exist, no corrupted data. Use automated data quality checks to scan for issues and report them. Agree before the upgrade begins what recovery option applies to the instance if the upgrade encounters problems, and who starts it. Document how long that recovery would take and how the team would reach the restored data. Rehearse each recovery step you have agreed before you upgrade, so you know how long it takes if you need it.

Test Upgrades in a Staging Environment That Mirrors Production

Upgrade a clone of your production instance in your staging environment first, before upgrading production. Create this staging environment several days before your planned production upgrade so you have time to test thoroughly. Test your customizations and integrations in the upgraded staging environment: do they still work, do they behave correctly, are there performance issues? Run your automated test suites against the upgraded system to verify critical functionality still works. This testing often identifies issues that can be fixed before production upgrade. If you discover a critical issue, you have time to resolve it or to delay the production upgrade. Document any issues you find and how you resolved them so your team knows what to watch for in production.

Coordinate Multi-System Upgrades and Dependency Order

Your ServiceNow system depends on other systems: connected applications, data sources, reporting tools. If those systems are upgraded at the same time as ServiceNow, integration issues are more likely. Coordinate upgrade sequencing with teams managing dependent systems. If System A feeds data to ServiceNow and ServiceNow feeds data to System B, upgrade in order: first System A, then ServiceNow, then System B. Document dependencies and the correct upgrade sequence. Plan communication with dependent teams so they know when upgrades are happening and can plan their own upgrade windows. Some dependencies might allow only one-way data flow; in that case, you have more flexibility in sequencing, but document the dependency anyway.

Automate Cutover and Recovery Steps for Rapid Failure Recovery

Upgrade Center schedules and monitors the upgrade itself, but your actions before and after upgrade should be automated where possible. Document your cutover procedure: when to stop business operations (if required), which systems to upgrade in what order, how to validate successful upgrade, and when to resume operations. If your upgrade fails and you need to run a recovery step, automate the steps you can run yourself so you do not waste time on them under pressure. Walk through the recovery steps before upgrade so you know how long they take and what the risks are. Communicate your upgrade window and rollback procedure to business stakeholders so they know what to expect and who to contact if problems occur.

Validate Performance and Configuration After Upgrade

After upgrade completes and you resume operations, monitor system performance closely for the first hours or days. Upgrade sometimes triggers performance issues that did not appear in testing: index changes, query optimization, memory requirements. Monitor database size, query performance, and user response times. If you see performance degradation, investigate whether it is related to the upgrade or a coincidental issue. Run data quality checks after upgrade to verify no data was corrupted during upgrade. Verify that your key integrations are still working correctly. Share post-upgrade status with stakeholders so they know the upgrade was successful and operations have resumed.

Choosing what to automate before choosing a tool

Automation in an upgrade program pays off when the task repeats every window, has a clear pass or fail result and fails in a way someone would otherwise miss. Tasks that meet all three conditions are good candidates. Tasks that depend on judgment, such as deciding whether a business risk is acceptable, should stay with people, with the automation supplying the evidence they decide on. Scoring each candidate against the criteria below keeps the first automation effort from going to something that looks impressive but is rarely used.

Candidate task Repeats each window Clear pass or fail Missed without it Automate now?
Checking that scheduled jobs ran in the last cycle Yes Yes Yes Yes
Comparing record counts between source and target Yes Yes Often Yes, once the tolerance is agreed
Judging whether a redesigned form reads well Sometimes No Yes No, keep it with a reviewer
Full visual regression of every page Yes Noisy Rarely Not as a first step

A build order for a regression suite that the team will keep

A suite fails when it is built all at once and then left to decay. Build it in layers, and keep each layer green before adding the next. The sequence below starts with the checks that protect the business most and leaves the broadest checks for last, when the team knows which ones it will maintain.

  1. Write the critical process list first, one line per process, with the business owner’s name beside each. The suite is only as good as the list it was built from.
  2. Automate one end-to-end path per critical process, using the Automated Test Framework or the team’s chosen tool, and run it on the test instance until it is stable across repeated runs.
  3. Add the integration checks next. Each check should send a known test message and verify the receipt on the other side, not merely that the endpoint answered.
  4. Add negative cases for the permissions that matter, such as a user who should not be able to approve their own request.
  5. Only then add broader coverage, and remove any test that has failed for reasons nobody has investigated in the last cycle.

The automated test framework guidance describes how these suites are structured inside the platform, and the note on governing automated testing covers who should own the suite once it exists.

Where automation creates more work than it removes

  • Browser-driven tests that rely on the position of a button rather than its label, which break on every cosmetic change and teach the team to ignore failures.
  • Scripts that copy production data into a test instance without a masking step, which move a privacy problem into a less controlled environment.
  • Cleanup scripts that run after a window and remove test data a later phase still needs.
  • Scripts that report success on the exit code alone, without checking that the expected records changed.

Checking that the automation is telling the truth

Before trusting an automated check, break the thing it is meant to catch on purpose, in the test instance, and confirm that the check fails. Keep a short list of these deliberate failures and re-run them after each release. Record who ran each deliberate break and when, so the evidence can be reviewed later. A check that has never been seen to fail gives little evidence on its own, and a dashboard built on such checks gives a false sense of coverage that is hard to correct later.

Frequently asked questions about upgrade automation

Should automation replace the manual upgrade test?

Not fully. Automation handles the repeatable checks, and a person still needs to use the process the way a real user would, particularly for forms and screens that were redesigned in the release.

How much of the suite should be run before each window?

All of the critical process paths and integration checks, on each run. Broader coverage can run on a schedule, so that the window itself is not the first time a broad test is exercised.

Who maintains the suite when the original builder moves on?

The release lead, with a named developer for each path. A suite with no named maintainer tends to decay, usually because a failing test is disabled rather than fixed.

The automated testing and release assurance page is a related reference for these suites.

Upgrade automation for ServiceNow: a go or no-go gate that automated checks feed

The most useful thing to automate in an upgrade program is the decision to proceed, not only the checks behind it. Each automated check writes a pass, fail or unknown result to one readiness summary, and a named person reads that summary and says go or no-go.

Without a gate, checks pile up as separate reports that nobody reads together, and the window starts because the calendar says so.

Set the rule before the results arrive

Agree the rule in advance, while nobody is under pressure. A workable pattern in most implementations looks like this:

  • Any failed critical-process test blocks the window until it is fixed or the business owner accepts the risk in writing.
  • Any check that did not run counts as a fail, not as a pass. Skipped tests are a common source of false confidence.
  • Failures in low-priority areas go on a fix-after list with an owner, and do not block the window.
  • A restore rehearsal that was not completed since the last major change to the environment blocks the window.

What the summary should show

Keep it to one screen: the date of the clone the tests ran against, the number of checks run, passed, failed and skipped, the open exceptions with their owners, and the decision with a name and time. If the clone is old, say so on the summary, because results from a stale copy describe a system that no longer exists.

Recovery is a decision too

Automate the steps of recovery so they run the same way every time, but keep the choice to start recovery with a person. Set the trigger conditions in advance, for example a failed login path or a stalled integration queue that has not cleared within an agreed time, and let the person on call confirm them. Slowness alone should not start recovery until someone has checked whether the upgrade is the cause, because an instance that is merely slow may be perfectly healthy.

The wider upgrades and release readiness guide covers the surrounding plan, and the testing and acceptance checklist is a good source for the checks that feed the gate. How changes move between instances is covered under update sets and release management.

Turning one scheduled-job check into a result the go or no-go gate can read

A scheduled-job check answers one question: did the jobs that should have run in the last cycle actually run on the cloned instance? Before the first run, write down three things: the jobs in scope, the expected interval for each, and the time window that counts as the last cycle. Without those, the check has no fail condition and will report a pass by default.

Take the job list from the business side, not from the platform. Ask which jobs feed a report, a sync or a control that a manager relies on. A job nobody depends on can stay out of the gate and go on the fix-after list. Each job on the critical list gets a named owner.

Run the check against the clone and read each job’s last run time against its window. Each job ends in one of three states:

  • Pass: the job ran inside the window and recorded no error.
  • Fail: the job did not run inside the window, or it ran and recorded an error.
  • Unknown: the job has no run history on the clone. For a critical job, count unknown as a fail, because a clone that never ran the job cannot show that it will work after the upgrade.

Two details catch teams out. The first is the clone date. A job whose last run predates the clone was never exercised on the copy, so it fails this window even if production looks healthy. The second is time zone. Compare every timestamp in one zone, write that zone on the summary, and a job that ran at the right hour will not look late by accident.

Keep the check narrow. It reads the run history and compares it with the window, and nothing more. Once a scheduled-job check starts judging what the job wrote, it has taken on a business decision that belongs with a reviewer. Whether the output is right is a question for the people who use it, and the business-owned validation approach is the better home for that. Record counts are a related check, and the reconciliation method used at cutover shows how to compare a source and a target without trusting one total.

Store each result as a row on the readiness summary with the job name, owner, window, last run time and state. That row lets a reviewer check the gate decision later.

Frequently asked questions

How long should we plan for a ServiceNow upgrade?

Depends on instance size and customization complexity. Small instances with minimal customizations upgrade faster; large instances with extensive customizations take longer, and the time varies with scope. Measure the duration in a test environment. Plan upgrade during a low-usage window so business impact is minimal.

What should we do if upgrade encounters an error and fails?

Most upgrade failures are temporary and can be retried. If retry fails, review the upgrade logs for specific errors, address the underlying issue (remove conflicting plugins, fix data quality issues), and retry. Only roll back if you cannot resolve the error.

Should we upgrade immediately when a new release is available?

No. Let other customers upgrade first and report issues. Wait at least a month before upgrading to a new major release so you benefit from others finding and reporting issues. Apply critical security patches immediately; apply other updates on a quarterly schedule.

Talk to Oblytech

Planning, running or fixing a ServiceNow programme? Tell us where you are and we will suggest a sensible next step.

Contact us