On Sat, May 24, 2008 at 10:25:16AM +0200, Marc Haber wrote: > > [...] > > > Is this approach flawed? > > > > Takes roughly twice the time. > > But only if recover is really needed which is the rare exception. > > What are we talking about, tens of seconds?
Yes, definitivly. Even a recover including a version upgrade on my systems take less than a second. In the recovery case the two perl subroutine calls take longer than the actual recover for an upgrade from libdb4.4 to libdb4.6 format. Of course, as measured on my system, and still being milliseconds an appropriate unit. > > > Just set the recover flag, as the underlying BerkleyDB function does > > the check whether any action is necessary itself and in the normal case > > returns very fast. > > So it never does an unnecessary recover? Yes, at least, according to the docs ;) Running it during normal operations just does nothing and therefore does not disturb running processes. The content of the db-file changes only when an old version is updated. Bye, Joerg
signature.asc
Description: Digital signature