Howard Chu wrote:
Robert Mertens wrote:
Hi Florent,
I am using the JDBM as a part of the Apache Directory Server. I am
switching from OpenLDAP because it's Berkeley did not always shut down
properly.
I find that quite difficult to believe. On a properly configured
installation we never see corruption or data loss. We've even tested
on machines that crashed due to failing DRAMs and the database
(containing over 1 terabyte of data) still recovered in under 2
minutes on the subsequent reboot (after replacement of bad memory).
Ever since BerkeleyDB 4.2 came out, we've had no data corruption
issues with BerkeleyDB...
I must say that I trust Howard on that ...
The JDBM seems to do the job, but does not seemt to handle
emergency shutdowns very well(or is it the Apache Directory Server?).
It depends on many factors. If you don't select a sync on write (ie,
differing writes for better performances), you may lose data. Otherwise,
a crashed JDBM base may need some recovery. We have to investigate this.
The BTree was corrupted when shutting down the machine it is running on
when we, at the same time, performed actions against the Directory
Server where we deleted entries. The result was that the Directory
Server could not be restarted again without replacing the JDBM. I am
waiting for Apache Directory Server team to implement a failover model
for such occasions.
This is something we have in our roadmap for 2.0.
--
--
cordialement, regards,
Emmanuel Lécharny
www.iktek.com
directory.apache.org
