BB credit transitions from zero

[joe08091977]

Hi all,

we have 8 MDS 9124 Switches. On all Ports connected to our storage systems (two HP EVA 4000)

i can see a high level of "bb credit transition from zero"

What does this means and what can cause this ?

Thanks for help !

[Kris Vandecruys]

Hi Joe,

This counter is the times that your BB Credits were 0 and freed up again.  This number is ridiculously high! This means your buffer is full most of the time and each time a spot frees up it's immediately filled again! This host is trying to drink from the metaphorical fire hose.

That said, BB Credits are the flow control mechanism in Fibre Channel. Since we don't drop frames in FC, we keep count of the buffers available on the receiving end of each link.  A port will not send any more frames if it knows the port on the other end has no more buffers to accept it. A buffer will be freed if the the receiving end can process it. For the case of the storage array, when it can actually perform an operation.

The most visible way you'll notice this is performance issues.

I'm working on a small video presentation explaining bb credits and slow draining devices, which I will post here once it's finished.

In the mean time, you can find some more info here: http://www.cisco.com/en/US/partner/docs/switches/datacenter/mds9000/sw/4_1/configuration/guides/cli_4_1/intf.html#wp1190054

[joe08091977]

Hi kvandecr,

thanks for the explanation.

Does it make sense to assign a higher number of bb credits for this ports ?

Thanks a lot !!

[kenk]

Hi Joe,

Adding additional buffers typically doesn't help.

It sounds like the end devices are on the edge of not being able to keep up.  I like to think

of the "transition from zero" counter as a performance warning track... well worth investigating

to determine if a device is oversubcribed.

Regards,

Ken

[Kris Vandecruys]

Hi Joe,

increasing would not improve your predicament, no, or only a little bit. But, like you noticed, this option isn't available on the smaller switches.  Don't worry tho, if you're not using very long distance links, the credits assigned by default are plenty for anything these switches will do.

The switches are probably not too 'small', but the attached devices could be processing traffic faster. Compared to the data they are receiving they are what is called 'slow draining'

The question you need to be asking yourself now is:  Is my environment impacted by this? If you're perfectly happy with the performance figures you are getting from your hosts, then this is not the first thing you should worry about.

There's also the chance that what you're seeing is perfectly normal.  Remember, this is a flow control mechanism.  If you have a higher-speed port consistently sending lots of data to a lower-speed port, that lower-speed port will show a high number of bb credit zero transitions.  It's just evidence that the lower-speed port is controlling the flow from the faster one. But 'lower speed' could be plenty fast for what you're running.