I've seen this occur when the growth setting is set too high, so the growth event takes too long. Obviously this depends on your storage environment and performance details.

In my case, the 10% default growth rate was fine when the database was small, but at some point when the database got huge enough that the growth event took over 30 seconds, which was long enough that the original query timed out and rolled back.

So check your growth rates, or just pre-size your tempdb to a larger size.

I've seen this occur when the file growth setting is set too high, so the growth event takes too long. Obviously this depends on your storage environment and performance details.

In my case, the 10% default growth rate was fine when the database was small, but at some point when the database got huge enough that the growth event took over 30 seconds, which was long enough that the original query timed out and rolled back. The rollback then cancelled the growth event, until the application re-tried the same query, and we got an endless loop.

So check your growth rates, or just pre-size your tempdb to a larger size.