ali.pouya@free.fr wrote:

Howard Chu wrote :

...

I see.

This has been fixed for back-bdb/hdb in HEAD. If the database was shut down cleanly before, there will be no scan on the next startup.

Hi Howard, Thanks for the corrections. We tested the HEAD extracted on monday november 3. Unfortunately the problem explained in point 2 persists (even after a clean shutdown). Our directory contains 10 million entries. It can evolve to 40 million entries next year. The local sacanning speed is around 250 000 entries per minute. So the scan of the recent objects on the standby mirror after startup can take several minutes (unless if we distribute the write operations on both mirrors).

I don't see this happening on my builds. Maybe you should post your configs and a sample of the data and modifications used to illustrate the problem.