We have recently converted a standalone SQL2000 server to ESX using the VMware Converter. The new VM appeared to be running well though we had an issue initially that forced us to disjoin and rejoin the domain. Once we started letting clients hit the database we started getting reports of loss of connectivity to the database from the enduser's application. We looked at the resources and the SQL server was running around 60-70% CPU utilization with 8GB of memory at 60% utilization. The disconnects are happening about every 20-50 minutes regularly and we have tried numerous things trying to rectify the situation including: changing the network adapter, patch cable, switch, etc. and of course the puzzling thing is that it worked well running off our 5 year old clone server, never a disconnect.
The new ESX machine is an HP DL380 G5, dual Quad cores XEON 3.0GHZ with 16GB of memory, 6 1000TX ports, Emulex HBA's connecting to an HP StorageWorks MSA1500cs SAN. We figured it would run well since the SQL server only shares the box with 1 other domain controller VM that is dedicated as a DC with light load. We even utilized CPU affinity to make sure the 4 CPU VM would have deidcated access to 4 of the XEON cores.
The developers are telling us that the SQL box is losing connection with the network and we verified that we are not able to ping the box during the 30-45 seconds of loss of service the clients are experiencing.
We swapped out everything on the network side including the NIC, the patch cable and the switch and even tried reconnecting the VM NIC to the port the original server connected to without issues for the last 4 years.
Are we missing something? Does SQL just have issues running on ESX in a VM and/or are their tuning parameters that we should be using to make it work.
Just hoping someone has seen this issue before and can shed some light on this for us. Thanks
Adam
