Latency between NFS and ESX host.. Not sure that I am seeing..

[StamfordRob]

All.

I have an ESX cluster of 3 hosts connecting to NFS share on DSS. I am seeing a lot of latency from the host to storage. When I look at the statistics the Bond shows 1000 packets received or transmitted averaged per hour.. When I look at some of the Eth via packets they show anywhere from 7.0 m received to 40.0m packets received per hour. Some of the Eth show as low as 140 packets per hour.

Bits per hour at the Bond are reading at 5m read. 5m write. BUT.out of my 8 ports bond three of the ports show 5m bits per hour combined. The other 5 ports average under 150 bits per hour both ways.

Below is my setup:

ESX host with 4GB 802.3ad bond to..
Cisco 6509 to...
DSS v6 server at 8GB 802.3ad bond..

From one of my ESX host reports we show 88ms for writes.. 30ms for reads.. the other two units are +/- the same.

Jumbo frames not enabled at this time. Is all of my traffic being directed to a specific set of ports in the bond?

What traffic numbers are just too high for a 1gb port?

Thanks in advance.

Rob

[To-M]

Try to go to the Console screen and enter CTRL ALT N then select the nics and then go to Info to see if there are any packet errors either for the RX or TX.

Also the 802.3ad should be same as it acts an 1 whole nic, also the test.log in ifconfig sections can show if nics have issues with packets and you might want to check the dmesg2 logs as well.