How long a migration takes
There is no honest single answer, because the work per page varies enormously: converting a text macro is instant, and replaying a draw.io diagram's full version history is not.
What you can do is measure. Run one space, read the pages-per-minute figure, and multiply.
#What makes it slower
Factor | Effect |
|---|---|
Behaviour version v1.1.0 | Replays every version of every draw.io and PlantUML diagram. Substantially slower on large estates. |
Behaviour version v1.2.0 | Also copies standalone app content: templates, saved searches, image libraries. |
Diagram-heavy content | Each diagram means fetching the source drawing, sometimes from another page. |
Rate limiting | Confluence throttles under load. The run backs off and retries rather than failing. |
A wide scope | The search reads far more pages than it changes, and reading still costs time. |
#What the progress header tells you
N of M pages checked, elapsed time, pages per minute, and an estimate to finish. The estimate is derived from the rate so far, so it settles down after the first few hundred pages and is unreliable before that.
#Why the tab has to stay open
The whole run happens in the browser tab. The run happens in your browser tab, so closing the tab stops it, and nothing notifies you when it finishes.
Check your power settings before a long run. A laptop that sleeps mid-migration stops the run exactly as if you had closed the tab. Everything written so far is kept; everything after it simply did not happen.
#A few things that catch people out
Measure on one space rather than estimating from page counts. The variance between spaces is large.
A run that stops part way has covered your oldest content, because pages are processed by ascending page id.
Download the log before doing anything else if a run ends unexpectedly. It is the only record of what was done.
#Related
Measure one space. Then multiply.
