Cisco MDS Slow connection

[mrmarie]

Hello,

We facing a latency issue for all LUNs that connected between Cisco MDS 9148S 16G switch and the storage, when I run command " show logging onboard error-stats" I see that counter "FCP_SW_CNTR_RX_WT_AVG_B2B_ZERO" is increasing in the two ports that connected to the storage, What is the problem?

Appreciate your help

Thanks

[Kirk J]

If you are seeing application latency issues, suggest you open a TAC case and have them look to see if you have slow drain issues, congestion, high TxWait, etc.

Sometimes issues at the host side, or storage solutions that are unable to keep up with IOPs demands can trigger the loss of B2B credits.

Fibre Channel/FC works on a buffer credit system: frames are only TX'd when it is known that the receiver has buffer space to receive. The TX and RX sides informs the other side on the number of buffer credits it has during initial FLOGI process.

Kirk...

[dlt113 · Accepted solution]

The port registering "FCP_SW_CNTR_RX_WT_AVG_B2B_ZERO" implies the attached device is not returning R_Rdy primitive to keep the Cisco MDS' fcx/y port from reaching 0 transmit buffers.   It can increment to some degree and still not be a problem.

That said, I suggest you first clear the fc counters on all switches involved in the path from host to storage then

when the issue is occurring capture a show tech details,  show onboard log and show tech slowdrain.  Then open

a tech support case with your SAN switch provider.   (Assuming your NX-OS is above 6.2.13)

Command line clearing of fc and ASIC counters:

clear counters interface all

debug system internal  clear-counters all