More theoretical at the moment.
But after reviewing the Exchange 2010 Content and Indexing blog.
I was wondering what kind of controls the Exchange Search Indexer would have to prevent a run-a-way Indexing storm if a user were to forward mail accidentally to an undeliverable address on another mail system?
The Event history table would I assume be updated everytime an NDR came in, which would trigger the Microsoft Search Indexer (MSExchangeSearch) to fire off the Microsoft Search Filter Daemon (msftefd.exe) to repeatedly parse the NDR and and pass that to
the Indexer (Msftesql-Exchange) service. (reminder: msftefd.exe is an "on-demand" spawned process not a service, so it doesn't appear throttled in any way)
Meanwhile the errant forwarding rule would forward the inbound NDR outbound again.. and so on Ad-Infinitude... or at least until all space in the users mailbox was consumed. With enough memory and cpu capacity I'd see the disk I/O being outstripped long
before the mailbox filled up. (and effectively perform a Denial Of Service to other users, or at the very least run msftefd.exe at a very high CPU percentage)
The "truly" scary scenario is the only documented throttling feature seems to be on the initial Crawl for a mailbox. But after that its carte-blanche access to system resources.
I'm sure Microsoft has thought about this potential scenario.. or someone else knows what would keep the system from plowing a black hole sized singularity in my mail server.
Eagerly waiting for someone to tell me this is nothing to worry about ;-).
-- John Willis, Facebook: John-Willis, Skype: john.willis7416