Hmph. No news processes running, but ovdb_stat thinks things are locked up.

Heath Kehoe heath.kehoe at
Thu Dec 21 16:34:34 UTC 2000

>Completely quiescent system that just froze up.
>(All news processes killed off).
>%ovdb_stat -k
>Lock Region Statistics
>            4000   Maximum number of locks possible
>               3   Lock modes
>               4   Current lockers
>            4000   Lock objects
>            4487   Lock conflicts
>        44191228   Lock requests
>        44191222   Lock releases
>               6   Outstanding locks
>               0   Deadlocks
>        51982226   Region locks granted without waiting
>           22074   Region locks granted after waiting
>          561152   Lock region size

BerkeleyDB implements locks using data structures in one of
the __db* files (which gets mmap'd by a process which opens
the database).  So when a process dies uncleanly, those
lock data structures survive the death of the process.  That's
why if something crashes and leaves a stale lock, it can
cause other processes to block in the database.

The recovery operation (performed by ovdb_init) will clean
up those stale locks.


More information about the inn-workers mailing list