I'm sorry if this is already fixed in 1.8, I have no easy way to test it but to upgrade my OS, but it might ring a bell somewhere (otherwise, just close the ticket).

On 1.4.6, when my postgresql backend (might apply to other databases) doesn't respond anymore (closed for maintenance, etc.), zabbix-server shuts down this way:

24260:20100113:025055 Connection to database 'zabbix' failed: server closed the connection unexpectedly
This probably means the server terminated abnormally
before or while processing the request.

24260:20100113:025055 Query::select oid from pg_type where typname = 'bytea'
24260:20100113:025055 Query failed:Result is NULL
24240:20100113:025107 One child process died. Exiting ...
24240:20100113:025110 ZABBIX Server stopped

It is particularly annoying, since I have no way of knowing if anything is down after!

I think zabbix-server should never shut down... Worst case, it should send email alerts saying something is wrong and retry.

Thanks for this great product.