output errors on HIFs port on fex2232P and TX pause frame are incremental

[mehtarahul1998]

Symtoms -iSCSCi traffic is dropping and Esxi's are getting disconnected frequently ,looked at the interfaces for individual host, I am seeing output errors and TX-pause frames increment 

BY default flow control is enabled  on TX and is disabled on RX ...Same I have been verifying the vmware team to provide the results of flow control configured at there side ? Any thoughts on this .

Thanks

Rahul Mehta

[Reza Sharifi]

TX pause accumulating on the interfaces is a common problem with 2232 series.  They come very little buffer (4mb) and so they run out of buffer with high speed server, storage, VMs, etc..

The only way you can remedy this is to go to a different FEX.  We had to replace ours with 2248PQ FEX.  They come with 32mb buffer and we don't see anymore pause frames.

HTH

[mehtarahul1998]

how to verify drops in buffers on fex ports ,Do I need to enter into CLI mode of fex and run the commands ,Also does enabling flow control on link level will help on this ...

[Reza Sharifi]

Enabling flow control does not really help. We tried it but had no luck. To see the buffer details you need go into CLI on the FEX or just open a TAC ticket and they will help you with that.

HTh

[Rajeshkumar Gatti]

Enabling flow control on the host side can help as it will respond to Tx pauses and stop sending traffic for a specific interval. Tx pauses are the outcome of the lack of buffers. There are commands to check for the drops on the fex but honestly it will not give you anything to help solve this problem.

Here are some of the common remediation steps

1)Maximise number of uplinks between  2232FEX and parent. This is to have ability to have more paths towards the network. Also it will help in having max buffers for N2H traffic.

2)If uplinks between FEX and parent are not evenly used, changing port-channel hashing can help. Reason is that if any of the Fex uplink gets more saturated then it leads to Tx pauses being sent towards the hosts port.

3)Since there is no local switching on FEX, avoid have east-west traffic flow profiles on hosts on FEX. 

4)Avoid having bursty appliances such as NAS devices, blade chassis on FEXes. These need to be on the parent.

Let me know if you need more clarification and I can help.

-Raj

[mehtarahul1998]

I have checked from host end and flow control is enabled on the ESXI host ...we are further troubleshooting on this...

ESXI and storage team are suspecting below remediation step for same ..

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1002598