We've been struggling with this for some time and have yet to find an answer.

All day long our BES server (v4.1 sp2) and our Exchange Server 2003 sp2 play happily together. No significant event log entries or bottlenecks.

However, as soon as the scheduled nightly maintenance begins, the BES server reports in its logs (look in the MAGT logs) that the exchange server is way too busy (blocked threads) and no messages get to or from the Blackberries.

At the same time OWA , Outlook Cached, and Outlook Online show no real performance issues. In performance monitor on Exchange, the disk, processor, and CPU figures do hop up but the server itself is responsive.

When the maintenance period (1am to 5am) ends, BES catches up and resumes normal operation.

Any ideas?

Related Question: Can regular maintenance be safely changed from nightly to weekends? Is 4 hours really required?

Exchange Server
Pentium 4 3.0 HT, 3gb ram, 120g of edb/stm data on SATA disks
Not many users at this time period but about 100 during the day.
There are 4 storage groups with 4 mailbox stores each with about 20 users

interesting question.. I'm going to check my logs for the same thing.

Hmm... is it possible that your database maintenance is affecting the SQL database that BES uses? Perhaps BES doesn't realize your maintenance is locking the database and interprets that as Exchange being unresponsive?

I know very little about network admin, so I'm just tossing this out to maybe stir the creative juices.

Thanks but it's not a possibility. The database maintenance period i'm referring to is in Exchange system manager on the 'Database' tab of the mailbox store properties. It's called 'maintenance interval' and it is defaulted to 1am to 5am daily - it peforms deleted mailbox and item purging in addition to internal defragging of the .edb .stm files. The BES database is a simple MSDE database on the BES server.

Update:
When we shrink the interval, the BES problem shrinks with it.

Also, the defragging by default can continue up to an hour beyond the maintenance interval. We successfully set the posted registry values to allow it only 5 minutes beyond the interval.

The server is getting major hardware upgrades in two weeks so we'll see if that fixes it.