[bithead]

As a network engineer I once had a project involving an infrastructure for a deployment of thousands (so I was told) of SQLServers. My part was pretty straightforward and ready well ahead of schedule, but the DB team got stuck because the SQLServer build process kept getting hung. The only response ever heard from microsoft was "It's a network problem"

So I dug into packet captures and found that the switches were using all-zeros gratuitous arps to populate the forwarding table more quickly, which is a fairly common practice. The SQL servers saw this as in indication of a duplicate IP address (the garps were for 0.0.0.0).

Technically this did trigger the duplicate IP address detection RFC, but was never a problem for other server builds or any other server operational capacities. Cisco was the other vendor involved and came out with an update to allow for a timing delay for the all-zeros GARP. However given the timeline runway and the need to do a bug scrub for just one access area in a large data center with other much more lucrative capacity builds the only fiscally responsible thing was to look to the vendor at the heart of the problem.

However it was not possible to get any response from microsoft on it in spite of having the most expensive support contract they offered. The monthly late fees for the project were in the 8 figure range, so microsoft was dumped since they were not answering the phone. It was not for a lack of trying either. Our so-called "high touch" liaisons were equally frustrated.

It was weird since we were not a small customer - we were one of their largest. If a fortune 10 can't get microsoft to answer the phone, what hope does anyone else have?