CISCO MDS - "transitions of BB credit out of zero state"

Hello

the B2B transition out of zero state message indicates that the B2B credits for a specific interface were at zero, and now are increasing again. This in itself is not abnormal, and indicates Flow control is happening on this interface.  The likely reason for this is that the interface is slower than another interface it is communicating with, ie a 4Gb interface sending to a 2Gb interface will cause b2b zero events because the 4Gb will fill up the buffer twice as fast as the 2Gb can drain it.  This is not an issue, just sub-optimal design.

If you're not experiencing worse-than expected performance, I wouldn't worry.

However please clear the counters and check if the values continues to increament

#Conf t

#interface fcx/y [ x: slot number, y:port number]

#clear counters interface fcx/y

#end

Check for

#show interface fcx/y counter details

Thanks

Hi Rangac

Thank you for the response.

The counters are definitely incrementing and this definitely seems in the right area. There are few more aspects to this....

The onward interface for communication is a 2Gb cross-site ISL for data replication (whereas 4Gbps source port), however that 2Gb interface has a large number of buffer credits (500) and is port-channelled with many other 2Gb interfaces to form a larger pipe. So will the flow-control 4Gb-> 2Gb still come into play in the same way?

The other aspect is that if it was due to a constraint on the 2Gb interface (that's part of the larger port channel), I'd expect to see similar results across all our other Symm ports hooking up in the same way, but we're not, seems to just be a certain one's

The final aspect is that this 4Gbps source interface is in 'shared' mode within the port-group, with potentially only 2.8Gbps bandwidth available.

Far from ideal, I know :-) , but just trying to establish if there is a bottle-neck here or not and what component before we put a plan in to address it.

p.s. The cross site 2Gbps links will be very difficult to change as they interface with a DWDM unit from service provider.

Cheers

Sorry forgot to add, the confusing matter form the original output is that the fact it's reporting...

"0 transitions to zero state"

Shouldn't mean we need to come out in the first place

I'd expect the counter to be >0, since each ISL is only 2 Gbps. So a single exchange is still only using a 2 Gbps link and not the whole portchannel. I'm facing the same thing right now: a dual 2 Gb ISL forming a 4 Gb portchannel and the distance is only something like 30 miles or so. I'm using MDS 9222i with 250 B2B credits on each side, on each of the interfaces and still I'm seeing hundreds of thousands "transitions of BB credit out of zero state". I didn't expect those at all. The only other explanation i can think of is that I have a "slow draining device" i.e. a storage array that's overloaded with data to be written, but the thing is that this VNX 5700 isn't really busy at all.