[f-nsp] Health Monitoring: TM Egress data errors detected

[Wilbur Smith]

Hello Folks,
The TM error counters you are seeing can be a bit cryptic. The TM (traffic
manager) is the chip that interfaces with the switch fabric (SFM) though
the SFM Links. When you see the errors on for Ingress or Egress, that is
relative to the chip, not the physical port. I the the reassembly errors
are the key in the output; the TM is having trouble putting back together
data that is originating from the backplane.

The MLX uses a CLOS fabric (every chip is exactly 2 hops for every
possible connection) so packets are split into “cells” as they are
forwarded across the backplane. When we see these error in the TM, it
usually means we have a hardware error between the TM and SFM. This could
be from a failing TM, a bad SERDES, or a bad SFM Link. In 5.4, the MLX
should be watching for any flapping SFM Links and attempting to
dynamically retune the SERDES to stop the error, but that won’t help if
you have a faulty TM.

Of course, an underlying code issue or bug could be causing these errors,
but I would try power cycling the module, reinserting it into the slot, or
making sure all the SFM Links and Fes (Fabric Elements in the SFM) are not
showing any problems.

If you’ve got support, I would check with Brocade TAC. They can run some
remote debug that will help isolate. Most of the output is in hex though,
so its a bit hard to cover here.

-WilburWilbur Smith
SE Ninja, Brocade 
wilbur.k.smith at gmail.com
wsmith at brocade.com


Disclosure: While I am a Brocade employee, my participation
in this community is a personal choice and not directed by my employer. And
information or recommendation I provide are my own and not an official
recommendation
from Brocade. Sorry folks, just need to make sure you know I’m ‘doin this
“off
the clock”!


On 11/7/13, 2:11 AM, "Franz Georg Köhler" <lists at openunix.de> wrote:

>Reassem Err Discard