After a hosting move, the app can open as usual while its overnight imports stop running.
Your migration plan for scheduled tasks has to cover the unattended application jobs that run while no one is signed in, such as imports, exports, reports, and background services. Those jobs don’t appear on a user list, so a move can pass all the daytime tests and still miss them. We’d rather find them in an inventory on Tuesday than hear about them from accounting on Wednesday.
Step 1: List the Imports, Exports, Reports, and Services
Start by writing down the night shift. This one is illustrative.
| Where to look | What may live there |
|---|---|
| Task Scheduler | Scripts, batch files, and program runs set up by your vendor or a past administrator. |
| Windows services | Background processes the application installs, such as import watchers and email handlers. |
| The database | SQL Server Agent jobs. Their metadata is stored in the msdb system database, outside the application database, so Microsoft says you must recreate them on the destination server instance, along with the logins, Agent service account permissions, and proxies they depend on. |
| The application itself | Its own scheduler or batch screens, which you see by opening the software. |
| Other machines | A workstation or another server that triggers the job from outside. |
Step 2: Record the Schedule, Identity, Paths, and Dependencies
Next, give each job a card, because five facts decide whether it survives the move.
| Record this | What can change after the move |
|---|---|
| Schedule | The time, frequency, and time zone. A new environment can use another time zone, and a daylight saving change can move a job. |
| Identity | The account the job runs as. Microsoft says password and S4U tasks need the Log on as a batch job right, and an S4U logon has no stored password and no access to the network or encrypted files. |
| Paths | Local drives, mapped drives, and network paths. Microsoft’s Ask the Performance Team recommends network (UNC) paths instead of mapped drives for scripts that copy files. |
| Upstream dependencies | Partner files, an email relay, bank or EDI connections, and license servers. Some partners restrict access to known IP addresses, so ask whether the new environment will look like a new address. |
| Output | Where results go, such as a folder, an email, or another system, and who reads them. |
Step 3: Check Whether Each Job Needs a Signed In User
A hosted application runs on a server, so a job that depended on someone being signed in at the old machine can stop.
| How the job runs | Needs someone signed in? | What to check |
|---|---|---|
| A Windows service | No. Services run in session 0 and can’t interact with a user’s desktop, per Microsoft. | Look for dialog boxes. A service waiting on a prompt the user can’t see can look hung. |
| Task Scheduler, with or without a logged on user | No. It runs without a signed in user and gets no interactive session. | Confirm the logon type, stored credentials, batch job right, and network access. If it fails, Microsoft suggests switching to the logged on user option as a test of the security context. |
| Task Scheduler, needs a logged on user | Yes. It runs when that user is signed in at the trigger time. | Check the host’s session time limits. Remote Desktop Services hosts can disconnect or log off idle, active, and disconnected sessions, and Parallels says to set those limits through Group Policy. |
| A script or macro that drives the application window | Yes. It needs a desktop to click on. | Ask your software vendor whether an unattended mode exists. |
Step 4: Test Output, Failures, and Repeated Runs
A job that starts isn’t a job that works, so run each one through six tests before the cutover. That matters most for overnight imports, because bad data spreads to other systems.
SET UP THE TEST FIRST
Use a test environment with a copy of the application and its database, copied input files, and test destinations. Make sure test emails, payments, and exports can’t reach production systems: use a test mailbox, a payment sandbox or no payment connection, and a test export folder.
| Test | How to run it | Pass when |
|---|---|---|
| Expected output, through the scheduler | Test the script by hand first, as Microsoft advises, to rule out script errors. Then start the configured task through Task Scheduler under its intended account, signed out if it runs unattended, and compare old and new results from the same copied input data. A hand run under your own account can hide permission and session problems. | Counts, totals, and file names match, and the task ran in its real security context. |
| Failure | Break one copied input, such as a missing file or a wrong password, and run the task. | The job fails in plain sight, alerts a named person, and leaves no partial data. |
| Overlapping run | Start the task again while the first run is going. A multiple instances policy decides the result: ignore the new instance (the default), run it in parallel, queue it, or stop the existing instance before starting the new one. | The policy behaves as chosen, and the data stays intact. |
| Repeated input | Let the first run finish, then give the job the same copied file again. | The repeat is skipped or rejected, and no records double post. |
| Missed run | Stop the test environment across a scheduled time, then start it again. | You know whether the job runs late, skips, or runs twice. |
| Timing | Compare the schedule with the environment’s clock, including a daylight saving change. | It runs when the business expects it to. |
The multiple instances policy controls overlapping runs but doesn’t stop the same file from being processed again after the first run finishes, so test both. Ask your vendor what the application does when the same file arrives twice.
Step 5: Assign Monitoring, Retry, and Cutover Responsibilities
Environment monitoring and job monitoring are different things, because a hosting provider can watch the servers but can’t know your import should have run unless someone agrees to watch it. So write each answer down before the cutover.
| Responsibility | The question to settle | Owner |
|---|---|---|
| Monitoring | Who checks that each job ran, and when each morning? | |
| Alerts | Who is told when a job fails, and how? | |
| Retry | Does the job retry on its own, and who decides on a manual rerun? | |
| Cutover | Who switches the old schedules off and the new ones on, and in what order? | |
| First week | Who watches the first nights, and who can fix a job at 6 AM? |
Microsoft’s guidance on scheduled tasks not running lists the signs to watch: a task that doesn’t start on time, stays in Running status, finishes without the expected actions, or shows errors in the History tab or Last Run Result. A clean status isn’t proof the import worked, so check the output. Ask the job’s author to add logging, and check the Task Scheduler Operational event log.
On cutover day, disable the old schedule before you enable the new one, so the job doesn’t run in both places. Freeze writes first, as our hosted application data migration checklist describes, and keep the old server until the first night’s results are reconciled.
Where CloudTop SaaS Fits
Bring your job inventory to the discovery call, because we map your application, users, integrations, data sources, and authentication, and verify integrations with your staff before go live. Monitoring the hosted environment is separate from monitoring your application’s imports and reports, so ask us what we watch and what we don’t. For legacy application hosting, our compatibility checklist and on premise migration guide cover the surrounding steps.
CloudTop SaaS is a CloudTop Office company, hosting Windows business software since 2000 through the parent brand.
FAQ
Want a second set of eyes on your job inventory?
A software fit consultation covers your imports, exports, reports, and services, and who tests and monitors them after the move.
