Meraki MS125-48FP and Intel Killer E2600 NIC

OVERKILL
Building a reputation

Meraki MS125-48FP and Intel Killer E2600 NIC

Had an interesting one today:

 

Swapping out an old stack of Catalyst 2960S switches for a pair of Meraki MS125-48FP's (linked at 10Gbit with SFP+) at a client's and was moving over their video surveillance stuff which includes an Acer Orion 3000 gaming PC as the video hub. Was working fine with the 2960S stack, however, when connected to the MS125's, the port starts to flap, drop down in speed and eventually just disconnects. 

 

Tried different cables, no change. Reconnect to the 2960, link is fine. 

 

Upgraded to the latest drivers from Intel, no change. 

 

As an interim measure I put a TrendNET unmanaged switch between the MS and the PC and that stops the issue from happening. 

 

Anyone else ever experience this? 

20 Replies 20
DarrenOC
Kind of a big deal
Kind of a big deal

What speed does that machine need to run at?  I’ve had issues with older devices wanting to run at 10Mbps.

Darren OConnor | doconnor@resalire.co.uk
https://www.linkedin.com/in/darrenoconnor/

I'm not an employee of Cisco/Meraki. My posts are based on Meraki best practice and what has worked for me in the field.
OVERKILL
Building a reputation

Probably 100Mbit, but it SHOULD work properly on Auto Negotiate. It initially links at Gig-E, stays that way for a couple of minutes, then drops down to 100Mbit, stays that way for a bit, then flaps some more, dropping to 10Mbit and then just flaps like crazy losing the link. 

 

Connect it to the 2960S, it links at Gig-E and is rock solid, same with the unmanaged TrendNET switch. 

PhilipDAth
Kind of a big deal
Kind of a big deal

If it is stepping down in speed it may be trying to use "Entergy Efficient Ethernet", which I am pretty sure the MS125 does not support.

 

Try disabling any kind of power saving for the NIC, or look for something called "EEE" and disable that.

OVERKILL
Building a reputation

Yeah, that was one of the first things I disabled, didn't make a difference. 

RWelch
Head in the Cloud

What firmware are the MS125s running?

If you found this post helpful, please give it Kudos. If my answer solves your problem please click Accept as Solution so others can benefit from it.
OVERKILL
Building a reputation

The latest, 16.9. 

RWelch
Head in the Cloud

Have you tried a different SFP+ module?

If you found this post helpful, please give it Kudos. If my answer solves your problem please click Accept as Solution so others can benefit from it.
OVERKILL
Building a reputation

The SFP+ is just the connection to the other MS125, the computer is connected directly via ethernet. I tried multiple ports and cables, same behaviour. As I noted in the OP, works fine on a 2960 or a cheap unmanaged switch. 

RWelch
Head in the Cloud

I am not familiar with this NIC but it might be worth checking the MTU size?

If you found this post helpful, please give it Kudos. If my answer solves your problem please click Accept as Solution so others can benefit from it.
OVERKILL
Building a reputation

MTU is 1500, should be fine. Jumbo frames are disabled. 

cmr
Kind of a big deal
Kind of a big deal

On another forum someone had a similar issue (but with a different vendor's switch) and resolved it by stopping and then disabling the Killer service.  If you can find a basic driver then that could also work.

If my answer solves your problem please click Accept as Solution so others can benefit from it.
OVERKILL
Building a reputation

Definitely can give that a shot. Bit thrown by the fact it works fine with the 2960 though, lol. 

cmr
Kind of a big deal
Kind of a big deal

Newer switches like the MS range try to talk to the client more than older switches like the 2960.  I've seen all sorts of problems with older or more esoteric devices and modern managed switches like failure to negotiate PoE, incorrect speed settings and repeated renegotiation issues.  Another set of problems is based around client NICs saving power and that could be an option to look into for you, see if the NIC has any power saving options and if so, turn them off. 

If my answer solves your problem please click Accept as Solution so others can benefit from it.
Brash
Kind of a big deal
Kind of a big deal

Yeah, I've seen some similar circumstances - especially for old devices and OT equipment.

Connect it to an old Cisco switch or a dumb switch and it works fine. Connect it to a brand-new Meraki switch and it has a fit and throws CRC's and/or speed negotiation fluctuations.

OVERKILL
Building a reputation

Though amusingly this is a quite new device, lol (gaming computer). And of course the NIC is from a rather ubiquitous vendor. 

cmr
Kind of a big deal
Kind of a big deal

Having done a little research, the NIC looks to actually be an old Realtek chip from 5-10 years ago that has been repurposed by Killer, as opposed to a native Intel part.  Not that it should make a difference...

If my answer solves your problem please click Accept as Solution so others can benefit from it.
PhilipDAth
Kind of a big deal
Kind of a big deal

PoE - interesting point.  The 2960S is probably not a PoE switch, while the MS125 is.

 

Have a go at disabling PoE on the MS125 port and see if that makes any difference.

OVERKILL
Building a reputation

Yes, the 2960S switches have 740W of PoE. I can definitely give that a go next time I'm on site though, just to see if it does anything. 

PhilipDAth
Kind of a big deal
Kind of a big deal

Hmm, ok, less likely then.

 

The 2960S would support pre-standard PoE, while the MS125 does not.

 

The straws are getting very short now.

OVERKILL
Building a reputation

For the sake of clarity, this is the old stack:

https://www.cisco.com/c/en/us/products/collateral/switches/catalyst-2960-s-series-switches/data_shee...

 

Specific model:

Cisco Catalyst 2960S-48FPS-L

 

Supports 802.3at PoE+

Get notified when there are additional replies to this discussion.
Welcome to the Meraki Community!
To start contributing, simply sign in with your Cisco account. If you don't yet have a Cisco account, you can sign up.
Labels