Øystein Grøvlen wrote:
David Sitsky wrote:
Some causes in the past are:
o DERBY-700
o DERBY-1838
o using durability=test mode
o running on hardware that doesn't actually sync disk when asked
o running separate jvm's on 2 machines accessing the same db across a
networked file system.
This is running on windows, we never set durability=test, so I highly
suspect the last point must have occurred at this site.
How about the second to last point? Most windows machines by default is
set up to cache writes in the disk controller. If the machine goes down
in an uncrontrolled manner, disk writes may get lost. Could this have
happened?
To be safe, one needs to turn off the disk caching.
The client informed me that the server hosting the data was always
shutdown in a controlled manner.
--
Cheers,
David
