Locking database cleanup?

[tridge at samba.org]

Jeremy,

 > Because a process can't catch or stop sigkill the smbd might
 > be in the process of allocating new space within the tdb, which
 > if it is killed might leave the tdb in an inconsistent state

Upgrading tdb in Samba3 to include the transaction code would fix
this, at last for databases where we can afford the slowdown that the
fsync() calls cause.

You can kill -9 a process within a tdb transaction at any time, and
you will either end up with the before or after transaction state, but
never in between.

Cheers, Tridge