10-14-2011 02:16 AM
The other day, I had an SFP which had gone bad, it turned out. There is a detailed description here: http://troelsarvin.blogspot.com/2011/10/what-dying-sfp-looks-like.html
This made me wonder: Would the graphs / error count picture had been different if it had been a cable problem instead of an SFP problem?
10-14-2011 04:15 AM
Perhaps not THE answer to your question but intresting nontheless is this blog posting >> http://www.thesanman.org/2011/01/crc-errors-code-violation-errors-class.html
Usually sfp problems can be detected by dropping in speed (if set on AN) and powerlevels. I'am using a figure of powerlevels in near 400 uW as "normal".
10-14-2011 04:50 AM
Thanks for the link - I'll read it tonight.
About power levels: Is there a way to monitor this? An snmpwalk doesn't seem to reveal anything power-related, but maybe I overlooked something.
10-14-2011 05:19 AM
From the FCMGMT mib you could try to work with loss of signal and loss of sync.
OID's are resp. .22.214.171.124.126.96.36.199.1.43 and .188.8.131.52.184.108.40.206.1.44
Haven't found OID's with regards to the sfp yet.
10-14-2011 05:29 AM
Found a Brocade document about DCFM in which the outline the possibility to monitor SFP powerlevels.
You'll need the fabric watch license for it to work.
If you got Fabric watch and configured it to your needs, you can only monitor on the event created when going through thresholds.
OID FW subtree is located on .220.127.116.11.4.1.1518.104.22.168.1.10