Slow NFS on VMware vSphere

[rockstoruser24]

I have Rockstor installed as a VM on SSD. I am running vSphere 6.0 and running on a 10 gbe NIC.

1. How do I confirm Rockstor (the OS) believes it can run at 10 gbps?
2. The NFS writes max out at about 500 mbps. Is there any way to troubleshoot?

[Spectre694]

1. You can use iperf to verify the raw speed of your network link

2)The quickest way would be to use something like to see roughly how fast the array is able to write.

dd if=/dev/zero of=/mnt2/{pool name}/testfile bs=100M count=10 conv=fdatasync

[Tim]

Well,I’m not sure which client is consuming the NFS endpoint. If you are using the NFS endpoint by vSphere itself (all-in-one setup?) than you have to keep in mind that ESXI is sending sync writes to the NFS share by default. Making it more secure, but slower.

Also see:

Rockstor VMware environment slow write speeds

So i have a Rockstor VM on esxi 5.5 host with pci passthrough asus pike 2008 raid card, 8 x 3tb seagate constellation es.2 drives. I have 2 issues one being the web interface for rockstor won’t populate S.M.A.R.T saying “Not supported” even though i can run smartctl commands all day with no issues. But i’ll probably make a separate post on that if no one has an easy fix, my main issue is with my write speeds. It seems to me what is happening is the machine is caching data into ram before writing…

[rockstoruser24]

[root@rockstor ~]# dd if=/dev/zero of=/mnt2/share/testfile bs=100M count=10 conv=fdatasync
10+0 records in
10+0 records out
1048576000 bytes (1.0 GB) copied, 3.13213 s, 335 MB/s

Looks like 2.6 gbps

[Spectre694]

Okay so the array is fairly quick so that’s good. I would look into making sure that NFS is in async mode (what @Tim mentioned)

I would also run iperf and see what that says the link is. There is a pretty good right up on it in this thread