I have a TCP socket server running as a Windows service running on a Windows 2008 server on a VMware complex at a cloud hosting provider. Sorry, but I do not know the details on exactly what VMware software or releases are there, but they at minimum have vMotion. In general, they are not on the latest release but probably a version behind.
Recently, my VM was "vMotioned" to another blade due to them starting some VM requesting 128G memory, and my software suffered SQL Server and TCP timeouts and SQL Server had a disk timeout. I was told the process took 10 seconds, but SQL Server complained that a disk I/O had timed out after at least 15 seconds.
So bascially, this "vMotioning" was not transparent to my service application, and I lost some connections, and had SQL Server errors. The VM and SQL Server seemed to survive just fine.
My question is - How can I code my software to better survive one of these vMotionings in the future?
I did increase various TCP and SQL Server timeouts already but are there additional guidelines and suggestions for software running in a vMotion environment?
Thanks, Dave