Problems sending e-mails to large numbers of recipients
Posted on 2014-01-21
We are running Exchange 2007, and our e-mail is set to use an external 'Smart host' to relay outbound messages - this is working fine in most instances.
However, some users send out e-mails to large numbers of users - maybe 200+ - using a combination of internal and external distribution lists - these e-mails are sent on a regular basis (ie multiple times per week) and we've been doing this without issue for years . The problem is that, whilst the internal recipients of these e-mails receive them OK, the external e-mails are not getting sent. Looking at the 'Queue Viewer' I can see the e-mail in the queue, with a 'Last Error' status of '421 Connection Dropped'
I first noticed this problem several months ago, but it seemed to coincide with busy times, and the e-mail would eventually go out. However, more recently it is happening with EVERY one of these large e-mails, and they don't eventually go - the sender will ultimately get a long list of undeliverable addresses.
I have a workaround which is to switch from using a 'Smart Host' to using 'DNS' for sending - this gets the e-mails out of Exchange, but is not really the answer to the problem.
I have spoken to the company who hosts our 'Smart Host' and it seems that the connection is being dropped because of a timeout. Apparently our server transfers across 100 e-mail addresses and then just stops transferring anything else - so after the timeout period expires, the connection is dropped. I'm not sure that 100 is necessarily a 'magic number', probably just a co-incidence.
This server (an HP DL380) also uses 2 * NC373i adapters, which I've seen mentioned in other discussions, suggesting that TOE and RSS should be switched off, as well as TCP Offloading. The NICs are teamed together, and appears that TOE is not enabled (as the TOE tab is not displayed in the team properties) and RSS is not enabled either. However, I do have 'IPv4 Checksum Offload', 'TCP Checksum Offload (IPv4)', and 'UDP Checksum Offload (IPv4)' all set to 'Rx / Tx Enable' and 'Large Send Offload Version 1 (IPv4)' enabled.
I have also noted that in the OS (Windows 2008 SP2), the command 'netsh int tcp show global' shows that 'Receive Side Scaling State' is 'Enabled' and 'Chimney Offload State' is 'disabled'.
Nothing on this server has changed since this used to work OK, apart from weekly installs of Windows updates (no discussion of pros/cons of this is required!). The drivers (including NICs) were all updated about 6 months ago, but it's worked fine since then, only recently getting to the stage where none of these e-mails will send.
Any ideas what may be causing this? Could it be the TCP Offloading? And if so, which of my settings should I change, and what adverse affects might this cause?
Or are there any other reasons why this might have started occurring?