Oh and I forgot to mention - no there's no real time av on either server

Hi all,

I had a problem on a Mercury server last weekend that took me quite a time to sort out.

What I eventually found out was that a user had forwarded their mailbox to another address but had in error put the address e.g. users.name.local.domain.

This caused an error on each core process poll after an incoming message for this user was received, sending a postmaster error message but the message file in the queue could not be deleted (notified in system messages). The number of queued messages built up with each core process cycle until there were 300+ at which stage Mercury crashed.

 I later reproduced the identical behaviour on my server.

I have 2 questions about this, why couldn't Mercury delete the message from the queue, and would it be a "good idea" to introduce some form of syntax check before the use of the forward file?

Cheers, Gordon 

I tried to reproduce this and got this result: I created a forward to <this.is.bad.com> and sent a test message to the user with that forward. The message was delivered, and on the next poll forwarding was attempted. This however failed with the console message "Recipient <this.is.bad.com> unknown". The address <this.is.bad.com> is presumably here considered a syntactically acceptable but nonexistent local address. Delivery failure notices are then generated to the postmaster account and to the sender, in this case the account that initiated the forwarding, thus causing a never-ending loop. Unlike your test all files in the queue directory were deleted after each pass, though, with no notifications in the system message console. The only thing that would cause a crash would presumably be when the disk eventually filled up. With multiple incoming messages the load of the system would of course increase as well.

Is there some real-time anti-virus program running on your server that might have prevented Mercury from deleting the queue files?

/Rolf

Hi Rolf,

Thanks for looking into this.

I wasn't looking at the server at the time it crashed, and I didn't take it that far with my server, but it definitely wasn't a disk space issue. I assume the crash is to do with the fact that all these files are locked i.e. I can only delete any of them by closing Mercury.

By the way, the server this occured on was Windows 2000 Server and my server is WS2008x64

I'll do some more testing on this in the next couple of days.

Cheers, Gordon