Mailing List Archive: nsp: force10

S50n w/ 10ge Input CRCs/Discards

Index | Next | Previous | View Threaded

doug at warner

Apr 28, 2011, 8:12 AM

Post #1 of 6 (1724 views)
Permalink

S50n w/ 10ge Input CRCs/Discards

We're trying to deploy 10ge upstream to our S50n stack (read, all the port 0
modules are stacking modules). We're seeing a decent amount of Input
CRCs/Discards and can't get the link to go > ~120Mbps. We've replaced
*everything* in this link (other than the upstream RSP) -- fiber, 10ge module,
XFPs and are still seeing this problem. Flow control is off on both ends; CPU
seems fairly low (<12%).

Any ideas on what might be going on here? Anyone else having problems w/ 10GE
on S50?

-Doug

Attachments:

signature.asc (0.19 KB)

marcus at grmpf

Apr 29, 2011, 10:09 AM

Post #2 of 6 (1634 views)
Permalink

Re: S50n w/ 10ge Input CRCs/Discards [In reply to]

On 28.04.11 17:12, Doug Warner wrote:
> We're trying to deploy 10ge upstream to our S50n stack (read, all the port 0
> modules are stacking modules). We're seeing a decent amount of Input
> CRCs/Discards and can't get the link to go > ~120Mbps. We've replaced
> *everything* in this link (other than the upstream RSP) -- fiber, 10ge module,
> XFPs and are still seeing this problem. Flow control is off on both ends; CPU
> seems fairly low (<12%).
>
> Any ideas on what might be going on here? Anyone else having problems w/ 10GE
> on S50?

We used to have 10GE on one S50n (no stack), but only moderate traffic
levels - around 1.2 Gbps. The other side was a S25p, no noticeable
problems/packet loss with this setup.

Marcus
_______________________________________________
force10-nsp mailing list
force10-nsp [at] puck
https://puck.nether.net/mailman/listinfo/force10-nsp

geekgirl at gmail

Apr 29, 2011, 11:54 AM

Post #3 of 6 (1646 views)
Permalink

Re: S50n w/ 10ge Input CRCs/Discards [In reply to]

Are you using LR optics and it's a short run? I've found f10 lr xfp's to be
finicky if they're receiving above -4db

Leslie

Sent from the road while on tour.
On Apr 29, 2011 10:53 AM, "Marcus Stoegbauer" <marcus [at] grmpf> wrote:
> On 28.04.11 17:12, Doug Warner wrote:
>> We're trying to deploy 10ge upstream to our S50n stack (read, all the
port 0
>> modules are stacking modules). We're seeing a decent amount of Input
>> CRCs/Discards and can't get the link to go > ~120Mbps. We've replaced
>> *everything* in this link (other than the upstream RSP) -- fiber, 10ge
module,
>> XFPs and are still seeing this problem. Flow control is off on both ends;
CPU
>> seems fairly low (<12%).
>>
>> Any ideas on what might be going on here? Anyone else having problems w/
10GE
>> on S50?
>
> We used to have 10GE on one S50n (no stack), but only moderate traffic
> levels - around 1.2 Gbps. The other side was a S25p, no noticeable
> problems/packet loss with this setup.
>
> Marcus
> _______________________________________________
> force10-nsp mailing list
> force10-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/force10-nsp

doug at warner

Apr 30, 2011, 6:50 AM

Post #4 of 6 (1635 views)
Permalink

Re: S50n w/ 10ge Input CRCs/Discards [In reply to]

We're using SR optics for a short run (~3m).

-Doug

On 04/29/2011 02:54 PM, Leslie wrote:
> Are you using LR optics and it's a short run? I've found f10 lr xfp's to be
> finicky if they're receiving above -4db
>
> Leslie
>
> Sent from the road while on tour.
>
> On Apr 29, 2011 10:53 AM, "Marcus Stoegbauer" <marcus [at] grmpf
> <mailto:marcus [at] grmpf>> wrote:
>> On 28.04.11 17:12, Doug Warner wrote:
>>> We're trying to deploy 10ge upstream to our S50n stack (read, all the port 0
>>> modules are stacking modules). We're seeing a decent amount of Input
>>> CRCs/Discards and can't get the link to go > ~120Mbps. We've replaced
>>> *everything* in this link (other than the upstream RSP) -- fiber, 10ge module,
>>> XFPs and are still seeing this problem. Flow control is off on both ends; CPU
>>> seems fairly low (<12%).
>>>
>>> Any ideas on what might be going on here? Anyone else having problems w/ 10GE
>>> on S50?
>>
>> We used to have 10GE on one S50n (no stack), but only moderate traffic
>> levels - around 1.2 Gbps. The other side was a S25p, no noticeable
>> problems/packet loss with this setup.
>>

Attachments:

signature.asc (0.19 KB)

joelja at bogus

May 8, 2011, 11:29 AM

Post #5 of 6 (1615 views)
Permalink

Re: S50n w/ 10ge Input CRCs/Discards [In reply to]

On 4/28/11 8:12 AM, Doug Warner wrote:
> We're trying to deploy 10ge upstream to our S50n stack (read, all the port 0
> modules are stacking modules). We're seeing a decent amount of Input
> CRCs/Discards and can't get the link to go > ~120Mbps. We've replaced
> *everything* in this link (other than the upstream RSP) -- fiber, 10ge module,
> XFPs and are still seeing this problem. Flow control is off on both ends; CPU
> seems fairly low (<12%).

how much of the traffic on the network is ethernet broadcast/multicast?

> Any ideas on what might be going on here? Anyone else having problems w/ 10GE
> on S50?
>
> -Doug
>
>
>
>
> _______________________________________________
> force10-nsp mailing list
> force10-nsp [at] puck
> https://puck.nether.net/mailman/listinfo/force10-nsp

_______________________________________________
force10-nsp mailing list
force10-nsp [at] puck
https://puck.nether.net/mailman/listinfo/force10-nsp

doug at warner

May 23, 2011, 12:26 PM

Post #6 of 6 (1554 views)
Permalink

Re: S50n w/ 10ge Input CRCs/Discards [In reply to]

On 04/28/2011 11:12 AM, Doug Warner wrote:
> We're trying to deploy 10ge upstream to our S50n stack (read, all the port 0
> modules are stacking modules). We're seeing a decent amount of Input
> CRCs/Discards and can't get the link to go > ~120Mbps. We've replaced
> *everything* in this link (other than the upstream RSP) -- fiber, 10ge module,
> XFPs and are still seeing this problem. Flow control is off on both ends; CPU
> seems fairly low (<12%).
>
> Any ideas on what might be going on here? Anyone else having problems w/ 10GE
> on S50?
>
> -Doug

We were finally able to get enough parts in to test between our S50n and an
S25n and were able to run the line clean.

After that we pushed things back to our upstream who finally tried us on a
different Supervisor card and we were able to run clean to them. It looks
like they had a bad 10GE port on their original Supervisor card.

-Doug

Attachments:

signature.asc (0.19 KB)

Index | Next | Previous | View Threaded

Interested in having your list archived? Contact Gossamer Threads