Use casePlatform replacement

A FileNet migration with no downtime

A service center works in FileNet six days a week. The last upgrade cost a weekend and two weeks of backlog. This time, agents must not notice.

Both platforms run side by side for weeksThe cutover is rehearsed with a stopwatchA person makes the go or no-go call

The operations engineBulk load, day 15 of 21

8.04 million documents loaded and checked while the agents work. Fewer workers by day, more at night, so FileNet stays responsive.

source · supervision, bulk load
Actions on this campaign
The verdict on the final batch

Both platforms agree · the lead can call go

01

The starting point

At a glance

The director of customer service set one condition: the agents must not notice the move. Eleven million documents cannot be copied in one weekend, and a platform used six days a week cannot be stopped for longer. A big-bang cutover is ruled out before it is planned.

For weeks, both platforms run side by side. The bulk load fills the new repository in the background, sync runs carry over the changes made in FileNet, and the switch only has to carry the last few hours. The cutover window is measured before it is booked.

Sync runsOperations engine
The gap, hour by hour. Each run carries the changes of the hour before, and its duration sizes the cutover.
Go or no-go, June 13Operations engine
Go, at 21:05. The criteria are written in the runbook. The call is Grace's.

Read the plan as four promises, kept in order: load in the background, keep the gap small, rehearse the switch, and freeze writes only for the time the last sync needs. Grace's team spends eight weeks on the first three and one Saturday evening on the last.

02

Copy while people work

FileNet stays the system of record until cutover, while the new repository fills in the background.

Step 1Planning

Write the rules both platforms live by

Grace Okafor, the migration lead, writes the coexistence plan for eight weeks. FileNet remains the only place the customer service application writes to. The operations engine copies into the new repository and keeps it in step. Nobody reads the new repository in production before cutover.

Coexistence planOperations engine
Old and new run side by side. Each batch is checked before the next one starts.
One system writes, the other followsThere is never a doubt about which platform holds the truth. Until cutover, it is FileNet.
Step 2Supervision

Run the bulk load around business hours

Daniel Weber, the integration administrator, starts the bulk load on 11.2 million documents with all their versions. To keep FileNet responsive for the agents, the batch runs with fewer workers during the day and more at night and on Sundays.

Mon to Sat, daytime: 4 workersNights: 12 workersSundays: 16 workers
Bulk loadOperations engine
The worker limits were set with the FileNet administrators, from their own monitoring.
The agents never feel the loadThe migration takes three weeks of background work, sized to leave the source room to serve its users.
Step 3Sync

Carry over what changed during the load

The bulk load took three weeks, and FileNet kept changing meanwhile. The first sync run carries over every change made since the load started: new documents, new versions, property updates. After that, each hourly run only covers the hour before.

Sync run S-001Operations engine
A change a control stops is listed. The run itself carries on.
The new repository catches up, then keeps upAfter the first run, the gap between the two platforms is one hour of changes.
03

Keep the gap small, then rehearse

For five weeks the sync runs every hour. The team watches the gap, then rehearses the switch end to end.

Step 4Supervision

Watch the runs shrink to a few minutes

Grace follows the sync runs on a timeline. Once the backlog is gone, each run carries a few thousand changes and takes a few minutes. That duration is the number that matters for cutover weekend: it says how long the final run will take once writes are frozen.

Sync runsOperations engine
The noon run carried a nightly import that had been moved to midday.
Has anyone measured how long your own platform would need to stay frozen for the switch? That number decides the weekend.
The freeze window is a measurementThe team knows how long the last run takes before it books the weekend.
Step 5Exceptions

Handle what a sync run cannot carry

A few changes need a person: a property value the mapping rejects, a name too long for the target, a content file missing in FileNet. Daniel works the exceptions list every morning, so none of them is still open on cutover weekend.

ExceptionsOperations engine
Each failed document is reprocessed on its own, without rerunning the sync.
Exceptions are cleared before the weekendCutover night is not the moment to discover a bad value. The list is worked down to zero in daylight.
Step 6Rehearsal

Rehearse the cutover in pre-production

Two weeks before the date, the team plays the whole cutover on the pre-production copy, with the real runbook and a stopwatch. Melissa Grant, the service desk lead, brings five agents on Sunday morning to work real scenarios on the switched application.

The rehearsal finds one problem: the letter templates still pointed at FileNet. A step is added to the runbook.

Cutover rehearsalOperations engine
Each step of the runbook timed, on a copy of the real platform.
The surprise happens on a rehearsalThe template problem was found on a Sunday in pre-production, with nobody waiting on the phone.
04

Cutover weekend

Saturday evening, with the call center closed, the team runs the rehearsed runbook plus its one fix.

Step 7Reconciliation

Freeze, final sync, reconcile

At 8:00 p.m. on Saturday the call center closes and Daniel freezes writes in the application. The final sync run carries over the last changes. The operations engine then produces the full reconciliation report: every document on its own line, with its FileNet identifier next to the new one.

Cutover runbook, June 13Operations engine
Fourteen documents were already broken in FileNet, and listed weeks ago.
The freeze lasts one eveningWrites stop for the time the rehearsal measured, while the call center is closed anyway.
Step 8Decision

Go, and switch the application

At 9:05 p.m., Grace runs the go or no-go call against the runbook's criteria: reconciliation clean, no open exception, template fix in place. All green. Daniel points the customer service application and the letter templates at the new repository, through its API.

Had any criterion failed, the answer was simple: no-go, writes reopen in FileNet on Monday, and nothing has changed for the agents. After a go, FileNet stays read-only until its switch-off date.

Go or no-goOperations engine
Until the call, FileNet is untouched. That is the way back.
The way back is decided before the switchThe team knew what no-go meant before Saturday. The call is a person's, with written criteria.
Step 9Viewer

Monday morning, agents at work

The call center opens at 8:00 a.m. on Monday. Melissa watches the service desk queue with the migration team on standby. Agents open customer files, send letters and add signed forms as on any Monday. Old letters open in the viewer, in the browser, as before.

Tickets about documents: 2One preview fixed by 09:40Hypercare closed at noon
LTR-2014-004417.pdfViewer
A letter from 2014, migrated in April, opened on Monday at 9:12 a.m.
The agents did not see the moveMonday looked like any Monday. The migration work happened in the weeks before and in one quiet evening.
05

The possible results

At 21:05 on cutover night, the final batch sits in one of three states, and each state gives Grace a different call to make.

Reconciled

Both platforms agree

Every document matches, the known source defects are listed, no exception is open.

Then: the lead calls go and the application is switched.
Exceptions listed

Some documents need a person

A few changes from the last hours were stopped by a control, each one named.

Then: fixed within the window, or the lead decides to wait a week.
Batch held

A control fails

Counts disagree, or a rule failed on many documents.

Then: no-go, so writes reopen in FileNet on Monday and the agents notice nothing.
What the operations engine does not decide

It does not book the weekend, freeze the writes or switch the application. It does not call go. Across the nine steps on this page, the cutover is decided by a named person, against criteria written in advance.

The engine loads, keeps the platforms in step and reconciles. The control is in the runbook and its rehearsal, so it holds on a Saturday night.

06

Time back

Measured on this FileNet exit and its eight weeks side by side. Your platform and your business hours will set your own.

0 business hourswith the platform frozen. Writes stopped after the call center closed.
1 rehearsalthat found the template problem before cutover night.
1 eveningfor the switch, sized on the measured duration of a sync run.

Where these numbers come from. They are counted on the scenario described on this page, with example data. They are not a customer result.

The rules behind the gains

Side by sideCoexistenceBoth platforms run together for weeks, and the old one stays the system of record until cutover.
Measure the windowSync runsThe final run is short because the hourly runs kept the gap small.
A person calls goRunbookThe criteria are written and rehearsed. The decision is the migration lead's.
07

The people involved

Replacing a platform people use all day needs the CIO, the administrator on call and the service desk lead to agree. Each one is protecting something:

IT and architecture

CIOs and migration leads

The business will not accept a weekend down, and the last upgrade proved them right.

  • Old and new run side by side during the transition.
  • The freeze window comes from a measurement.
  • Until go, the source is untouched.
  • FileNet, Documentum, OpenText or SharePoint sources.
Administration

Integration administrators

On Saturday night, you are the one on the phone.

  • Fewer workers by day, more at night, to spare the source.
  • Failed documents are reprocessed without restarting.
  • An interrupted batch resumes from its checkpoint.
  • Throughput, errors and remaining volume, while it runs.
Business operations

Service desk leads

Monday morning is yours to answer for, whatever happened on Saturday.

  • Agents test the switched application on a rehearsal.
  • The cutover happens while the center is closed.
  • Old documents open in the viewer as before.
  • A no-go leaves the agents on FileNet, unchanged.

A good first step

With one week of changes. Run the bulk load of one class into a test space, then a week of sync runs, and measure how long each run takes. That number sizes your cutover.

FileNet stays the system of record the whole time, so the agents work as usual while you measure.

Next step

Size your own freeze window

Tell us how your platform is used across the week. We will plan a test load and a week of sync runs with your team, and measure how long your cutover would need.

The screens on this page are drawn to show one step at a time, with example data.

Uxopian Software helps organizations put AI to work across the systems they already run, without giving up control of their data, their infrastructure or their critical processes.