GBIC moved from Cat 3508 --> Cat 6513 shows wrong media type

Greetings

This weekend I tried to move a fiber connection from one Catalyst 3508
to our central Catalyst 6513 equipped with a 16 port GBIC module
(WS-X6516A-GBIC). What I thought would work was to remove the GBIC
(G5486 LH/LX) along with the fiber connection and mount the GBIC in an
available slot in the 6513. However the media type of the GIBC is now
shown as LH instead of LX as was present in the 3508 and the link is
shown as down/down
I cannot find any indication that this can be controlled using console
commands on the 6513. Has I missed something or do I have to get a new
GBIC? As far as I know the other end of the fiber connection is out of
my control.
Our IOS is version 12.2(18)SXF3

Regards
Fredrik Hofgren

 

LX/LH is the same thing. Reverse the fibers and see if you get link.

 

If he moved the GBIC and fibres together then that wouldn't be a
problem. Is a 3508 old enough not to do autonegotiation? If so he may
have to configure the port on the 6513 - "speed nonegotiate" is the
interface command, I think.

Sam

 

Fiber GBICs don't negotiate, at least I have never seen one that does,
please correct me if I'm wrong, supporting documentation as well please.
There are 2 different styles of GBICs, the original which had little
release levers on the side, those you did not need to remove the fiber to
remove the GBIC. Then there were these cheesy POS ones that had this stupid
little wire type bar that you had to lift to remove them, on those you
needed to remove the fibers.
A GBIC is a GBIC is a GBIC, they work in all chassis, there is nothing
platform/model specific about them. Never seen one fail (not saying they
don't <G>). There are only a few things that can prevent them from coming
up.
1, They need to be reversed.
2, The port is admin down (but then it would be amber).
3, The port is error disabled (never seen a fiber error disable, could be
disabled if bpdugaurd was on).
4, Bad fiber, doubtful where this worked before, but I suppose the patch
cable could have been damaged when it was moved.

 

Golly, no! The GigE standard says that you should always negotiate and
if you fail then the interface doesn't come up. That's to prevent the
duplex mismatches that plague 10/100M ethernet. Prestandard products,
such as the Cisco 12000 family GigE boards don't negotiate so if you
want to connect them to a 6500 then you have to turn off negotiation on
the 6500. You then have to remember to turn negotiation back on again
if you later connect that same 6500 interface to a standard device.

I don't have a reference to the standard but this is a clue:

<http://www.cisco.com/en/US/products/hw/switches/ps708/products_command_r
eference_chapter09186a00801eaed9.html#wp1022290>

The second type is an SFP (a Small Form factor Package GBIC, GBIC stands
for something like GigaBit Interface Connector) and is much more
commonly referred to that way.

Not true. Some Cisco platforms don't (or didn't) support at least one
version of the Cisco 1000base-T GBIC (it takes a higher current than the
optical ones). GBICs from different manufacturers cannot be reliably
mixed (largely due to FUD, it seems). For instance a 3Com GBIC will
probably work in a Cisco chassis but the type will not be reported
correctly (it's always been blank when I've seen it) and a Cisco GBIC
simply won't work in any of the 3Com switches we've tried them in
(though 3Com have such a bewildering array of kit that I couldn't rule
out them working in some). Here's some info:

<http://www.cisco.com/en/US/products/hw/modules/ps5455/products_device_su
pport_table09186a0080446625.html>

Cisco GigaStack GBICs and other things are different again. There's a
product overview page here:

<http://www.cisco.com/en/US/products/hw/modules/ps5455/products_device_su
pport_tables_list.html>

Here's Cisco's policy on non-Cisco GBICs:

------------------------------------------------------------------------
Note: Cisco-approved GBIC modules have a serial EEPROM that contains
the module serial number, the vendor name and ID, a unique security
code, and cyclic redundancy check (CRC). When a GBIC is inserted in the
switch, the switch software reads the EEPROM to check the serial number,
vendor name and ID, and recompute the security code and CRC. If the
serial number, the vendor name or ID, security code, or CRC is invalid,
the switch places the interface in an error-disabled state.
------------------------------------------------------------------------
------------------------------------------------------------------------
Note: If you are using a non-Cisco approved GBIC module, remove the
GBIC module from the switch, and replace it with a Cisco-approved
module.
------------------------------------------------------------------------

from
<http://www.cisco.com/en/US/products/hw/switches/ps646/products_installat
ion_guide_chapter09186a00800a92f5.html>.

They do.

5, check negotiation (see above)
6, clean the fibre ends (very important for single mode cabling).

Sam

 

Errr, no.

There are two types of GBIC as described. Brian is not using
the term GBIC to mean an arbitrary interface adaptor module thingy.
He means a GBIC.

I have heard the word GBIC used as a term for an arbitrary
interface adaptor module thingy so I understand
your reasoning.

 

OK, I've never seen a traditional GBIC with a wire latch.

SFPs definitely have them, and in several places [1] SFPs are referred
to as "Small Form-Factor Pluggable (SFP) GBIC Module" (yes, Pluggable
not Package - sorry). So SFPs *are* GBICs with a "stupid little wire
type bar that you [have] to lift to remove them".

Sam

[1] here are a few - Google for "SFP GBIC" to find lots more:
<http://www.cisco.com/univercd/cc/td/doc/product/core/7200vx/72vxfru/5067
g.htm>
<http://www.nanog.org/mtg-0310/pdf/wodelet.pdf#search="SFP GBIC">
<http://www.netgear.co.uk/fibregigabit_module_agm731f.php>

 

This is an interesting topic/conversation to me, please don't take
anything I'm saying wrong. I would love to be proven wrong on this and find
a way to bring up legacy customers shy of using media converters.
I do not believe the FIBER GBICs (we are talking what the OP has,
5484's, 5486's or 5487's) follow the GigE standard, these might have even
been before the standard. Again, I believe that their speed/duplex cannot be
changed at all. They are 1000, that's it, plain and simple. I have done
numerous searches on them and all I see mentioned is 1000base. I also just
tried this is a lab, going from a 3550 with a 5484gbic to an old 2924M with
a 100M fiber port, the link will not come up. While this sure as heck isn't
definative, it does support my thinking. You can google groups on people
trying to bring up new switches to their older devices which were only 100M.
One thingthat I am finding which supports your statements is that they do
support the speed nonegotiate command, but my thinking here is that it is
nonnegotiate for flow control, not speed as indicated by the "speed" in the
actual command. Another supporting tidbit is the fact that there is no
duplex settings to manipulate, if it did follow the standards that should be
a configurable parameter as well.

Again, was refering to what the OP has which are 5484's, 5486's or 5487's.
He came from a 3508 there are no other fiber GBIC's available for it.

Was refering to yet again what the OP has, not SFP's, not GigsStacks. They
are Cisco GBIC's, not 3com, extreme foundry or anyone else. They are not
platform independant and are a standard accross all of their platforms that
accept that size/model GBIC.

I'm sure they do! LOL

 

I was strictly refering to what the OP had. There are tons of different
model GBIC's out there. I should have been more clearer (great english huh?
LOL).

I can shoot ya a pic if you'd like! I hate these friggin things! <G> Got a
few of them here in my lab. I'll trade you for some of the other ones so you

 

I worked with GBICs extensively about 7 years ago and at
that time there were quite a few failures.


As to the autonegotiation issue. As far as I know Fiber Optic
Ethernet ports each only offer a single speed.

In the 10 and 100 days this was because the different standards
specified quite different wavelengths and the physical optics
and devices had to be different.

I don't know how GBE is vs the others though.

Constraining a port to a single speed though does not
prevent it Autonegotiating. That is exchanging mesages about
capability with its neighbour.

In the Fibre Optic GBE case they will both say 1000 FD
and then they will happily go to work.

comp.dcom.lans.ethernet may have something
or if you ask a question there Mr Seifert may be around
or you could buy his book/get the IEEE Standard.

My guess is that the Standards committee may have made
Autonegotiation compulsory even though in some cases
it is never going to be needed in order to simplify
the Standard. I don't, I hasten to add, know.

It is in the same veign interesting to note that the GBE standard
included Half Duplex operation as a possibility however
I believe that HD implementations are non-existent.
I suspect that the Cisco Stacking GBIC is HD but it is intended to
interoperate only with others of its kind and as such may
not follow any standards at all.

 

OK, let me clarify. The negotiation is done by the GigE interface logic
and is basically independent of the kind of GBIC or other PHY installed.
I agree that most, if not all, GBIC based GigE systems don't negotiate
speed, but they do negotiate duplex and flow control. (It turns out
that 1000base-T systems cannot operate without negotiation turned on;
other 1000base-X systems have it on be default but are allowed by the
standard to have it configured off.) Here's a description for CatOS but
there ought to similar things in the IOS manuals:

<http://www.cisco.com/univercd/cc/td/doc/product/lan/cat5000/rel_4_4/conf
ig/ether.htm#xtocid2934713>

In the example we have here there is a 6500 with a standard Cisco
1000base-SX GBIC (that'll be a 5484) connecting to a Cisco 12000 1-port
GigE interface. That also has a GBIC but according to the C12000 data
sheets it seems to be a special part. No change of speed (or duplex)
but we do have to turn of negotiation to make the link come up.

Yes - agree. Apologies if I misled you.

Yes. From the URL above:

"With Gigabit Ethernet ports, autonegotiation is used to exchange
... duplex information (even though Cisco Gigabit Ethernet ports
only support full-duplex mode)."

OK, the second type isn't an SFP. But none of the 54xx series has
the wire latch, does it?

I missed the OP saying 5484/6/7. Those do seem to be interchangeable
(see that URL above) but there are lots of others that Cisco make or
made that aren't, so in general a GBIC, even a Cisco one, is not a GBIC
is not a GBIC.


Sam

 

OK, feel free to send a photo - I'll keep an eye on my spam traps. I'll
decline the offer of a swap, if you don't mind.

Sam

 

Email sent. Let me know if you don't recieve them, I can put them in an FTP
for you. Are you sure you don't want to trade? <G>

 

Another correspondent has just sent me this:

<http://www.cisco.com/en/US/products/hw/routers/ps368/products_installati
on_guide_chapter09186a008069bab8.html#wp1041652>

That's for 7600s. The apparently corresponding page for 6500s

<http://www.cisco.com/en/US/products/hw/switches/ps708/products_installat
ion_guide_chapter09186a008020e0ae.html#wp1077095>

only shows the non-wire-latch types. Duh...

Sam

 

3, The port is error disabled (never seen a fiber error disable, could be

I have, caused by a link flap, fixed with "no errdisable detect cause
link-flap"

James