I'm using the GhettoVCB script to backup two VMs from ESXi 3.5 to an NFS share created by a Windows 2k3 server connected via Gigabit ethernet.
The problem I've run into is that these backups are excruciatingly slow. An 8GB backup took around 4-5 hours, and a 200gb VM which started at 3am is only 10% done as of 11:30. That gives me a rate of about 2GB per hour. (Assuming I did my math right, 582KB/sec, or 4.660 Mbps) I've searched for others with similar issues using Google, but I'm not sure if the issue is with the NFS server on Windows 2k3, or if the problem is with ESXi, or some settings in-between. Also, I noticed that the NFS mount I set up by following the instructions on this blog is no longer showing up in the VMIC, but is still mounted on the ESXi server. However, even the first backup was still slow, even when it still showed up.
you should check for errors on the NIC and on the switch. Make sure duplex etc are setup correctly on the NIC and the switch.
david
A simple IOMEGA $100 dollar NFS server can give you 20GB per hour at least so I would say you have a problem.
The switch is a Netgear PROSAFE® 24-PORT GIGABIT ETHERNET SWITCH (Model #JGS524). Both machines are plugged directly into this switch. Indicator lights show that they are indeed connecting at gigabit speeds.
On the source: In VMIC, Actual Speed, Duplex indicates 1000Mb, Full Duplex for the ESXI host
On the destination: In Device Manager, under Advanced Settings, Speed/duplex settings is set to "Full auto-negotiation". And Local Area Connection status indicates speed is 1.0 Gbps.
I don't know how to check NFS duplex settings on either machine (is there such a thing?)
I'm using the GhettoVCB script to backup two VMs from ESXi 3.5 to an NFS share created by a Windows 2k3 server connected via Gigabit ethernet.
I've never used NFS in M$-environment, but can you set the share to async mode? It made a huge difference for me, and should be quite alright for a backup server.
async This option allows the NFS server to violate the NFS protocol and reply to requests before any changes made by that request have been
committed to stable storage (e.g. disc drive).
Using this option usually improves performance, but at the cost that an unclean server restart (i.e. a crash) can cause data to be
lost or corrupted.
</quote>
I get sustained speeds of >100MByte/s to my linux NFS share when running GhettoVCB over my gigabit network.