It looks like I have run into my first real issue with VMware Server. After many successes I have finally come across a problem. We are ramping up for some testing for an ERP upgrade. To help allow us more complete testing I decided to role our current physical test ERP server into a VM. Day to Day operations in that test VM have been working fine. I have done some DB restores, people have been using the VM for normal testing and no issues to report.
Last night we tried our first upgrade of the software. It failed part way in saying it had lost connection to the database. We looked at the errors just a bit, but decided to roll back to our shapshot and then re-run it. No point wasting time on what could have been a fluke. The second upgrade also failed. There were several errors including ones saying we needed to run DBCC CHECKDB on the database. We did that and it returned clean, no errors. Running it again though resulted in a lost connection with the server error. More digging in the SQL logs also showed I/O timeout errors as well.
Doing some searching through the VMware forums show there are several others that have had this problem. So far, no one has a good solution to the issue. The best I have seen so far is to create a VM in GSX Server and then upgrade to VMware Server. I would rather avoid doing that, so I have not tried it yet.
In looking at the VM it appears it is using the buslogic driver (the OS is Windows Server 2003). That is not the ideal driver, so I will try switching that to the LSI driver tomorrow and also increase the amount of RAM available to the SQL VM to see if that can help us get by the problem. If not it looks like I will have to go back to a physical box for a bit to keep from holding testing up at this early stage.
Seems so odd that I am seeing this issue though. I did some serious testing in VMs with Exchange upgrades, restores, etc and had no issues at all. Some of the Exchange restores included 15GB restores, much larger than the DBs I am working with now. Hopefully I will get this figured out or VMware will come up with something – it doesn’t appear I am the only one facing this issue.