enc_out and bad_os errors on one side of an long distance IFL

[Anonymous]

I have a Brocade 48K with a routing blade, connected across single mode fiber to a Brocade 300 switch on the other side.  This connection is less than 10KM.  We have established the IFL, however my end is seeing constant enc_out and bad_os errors, and we have not put any traffic across this link yet.  The 300 is not seeing any errors on the port.  I have reseated the jumper and changed out the optics on my end, reset the port, and cleared the stats.  Still growing enc_out and bad_os errors.

Any clues from the experts out there on how to troubleshoot this?

[felipon]

Hi,

Could you paste the output of portcfgshow of both ends?

Is auto-negotiation enabled? Disable it. What is the current link speed?

Rgds

[Anonymous]

The speed is 4G

From the 300:

Area Number:              15
Speed Level:              AUTO(HW)
Fill Word(On Active)      3(A-A then SW I-A)
Fill Word(Current)        3(A-A then SW I-A)
AL_PA Offset 13:          OFF
Trunk Port                ON
Long Distance             LE
VC Link Init              ON
Locked L_Port             OFF
Locked G_Port             OFF
Disabled E_Port           OFF
Locked E_Port             OFF
ISL R_RDY Mode            OFF
RSCN Suppressed           OFF
Persistent Disable        OFF
LOS TOV enable            OFF
NPIV capability           ON
QOS E_Port                OFF
Port Auto Disable:        OFF
Rate Limit                OFF
Mirror Port               OFF
Credit Recovery           ON
F_Port Buffers            OFF
Fault Delay:              0(R_A_TOV)
NPIV PP Limit:            126
CSCTL mode:               OFF

From the 48K:

Area Number:	76
Speed Level:	AUTO
AL_PA Offset 13:	OFF
Trunk Port	ON
Long Distance	LE
VC Link Init	ON
Locked L_Port	OFF
Locked G_Port	OFF
Disabled E_Port	OFF
ISL R_RDY Mode	OFF
RSCN Suppressed	OFF
Persistent Disable	OFF
NPIV capability	ON
QOS E_Port	ON
EX Port	ON
Mirror Port	OFF
FC Fastwrite	OFF

I'm assuming that the problem is the 'fillword' being arbff/arbff.  I will have the customer set the portspeed to 4G.  Should I also have them fix the fillword?

[felipon]

Hi,

Fillword shouldn't care at 4G, but I would set it to 0 in both ends, just in case... also remove auto-negotiation out of the equation. (try setting the speed to 2G, to see it this behavior remains)

enc_out can be related to physical layer issue, check the Tx and Rx Power in both ends (paste the output of sfpshow <slot/port> -f)

What type of fiber cable and SFP are you using and how many patch-panels (if any)?

Do you have more IFLs between these switchs?

Sorry for making so many questions, it is just to figure out what could be going on.

Rgds

[Anonymous]

No problem and thanks for your assistance.  This is definitely related to the IDLE/IDLE negotiation, and probably the low level of the FOS installed on our 48K.  After a stats clear and a port bounce, the enc_out and bad_os counters would increment at a fast pace with NO frames being passed.  Locking the ports in at 4Gb and bouncing the ports at both ends did NOT help the problem.  What we had to do was set portcfgfillmode p# 0  and before that would work, we had to make vc_init=0 on both sides (not an ideal situation).  The link is pretty stable now with normal performance.  I plan to upgrade the FOS on the director soon, to help with the compatibility issues (hopefully).