[c-nsp] best practice for parity errors in NMS

[Peter Rathlev]

On Fri, 2015-04-24 at 10:52 +0200, selamat pagi wrote:
> In case of most parity errors, Cisco recommends to:
> ...Monitor the system regularly for reoccurrence. If no further events are
> observed, it is a soft error. If the error occurs frequently, request a RMA
> ...
...
> What would be a good number ? How to you handle this on your NMS ?

I wouldn't be comfortable with any more than a few each year from any
one device. At most.

Soft errors are caused by interference (e.g. solar flares) and if your
devices experience them often then either the hardware if faulty or you
should look at some more screening from interference.

We handle these in a "laissez faire" way; we have few enough devices and
are few enough people for us to generally remember what happens. If we
start seeing what we think are too many messages (e.g. more than one in
some weeks) we pull statistics from the log archive.

We currently have a single device (Sup720-10G) that logs an error about
once or twice a year. It sits just beneath a box with almost the exact
same configuration and that one doesn't log any errors. But since they
are (so far) recoverable and rare we have done nothing yet.

(Improperly seated cards might also be a cause for errors by the way.)

-- 
Peter