Sequencing a large estate
Above a few thousand pages, a migration stops being one action and becomes a plan. The tool does not sequence for you, so the sequencing is done with search filters.
The good news is that re-running is safe and incremental, so working through an estate in pieces costs nothing extra.
#A workable order
Wave | Scope | Why |
|---|---|---|
1 | One pilot space | Learn what your content does |
2 | Two or three spaces with a known owner | Prove the review process, not just the tool |
3 | Everything else, space by space or by label | Predictable, reviewable, and easy to pause |
4 | A final site-wide run with no filter | Catches the stragglers nothing else selected |
#Why space by space
Each run is short enough to sit through, which matters because the tab has to stay open.
Each run has an owner who can review it.
A problem is contained to one space rather than the whole site.
The log per run is small enough to be useful for repair and rollback.
Pages are processed oldest-first by page id, not by importance. A run that stops half way has covered your oldest content. That is another argument for short, scoped runs.
#The final sweep
Finish with an unfiltered run. It is fast, because everything already migrated drops out of the search by itself, and it catches the pages no filter happened to select.
#A few things that catch people out
There is no re-run-failures action, so it is worth labelling pages that fail and filtering on the label for the next attempt.
Rate limiting is handled by backing off and retrying, so a big run gets slower rather than failing.
Keep every run's log. Repair and rollback both need the log from the specific run that caused the problem.
#Related
Space by space. Finish with a sweep.
