BladeSystem - General
cancel
Showing results for 
Search instead for 
Did you mean: 

VCM fail over question.

The Brit
Honored Contributor

VCM fail over question.

I currently have c7000 enclosures, and I have ENet VC modules in interconnect slots 1, 2, 3 and 4. CX4 stacking links are in place between 1 - 3, and 2 - 4.

Under normal circumstances the VCM runs in IC module 1, and in the event of a failure, it will move to IC module 2.

My questions (hypothetical if you wish) are;

1. If IC module 2 also fails, will VMC move to module 3?

2. If a ENet module fails, is there still a path through the failed module, say from 3 -> 2?

3. (slightly different topic), if we switch to a 10GB network and use the CX4 cables as uplinks, can the 8x1G ports be combined and configured as stacking links between modules?

Thanks for your help

Dave.
5 REPLIES
JKytsi
Honored Contributor

Re: VCM fail over question.


1. My (hypothetical) answer is that bay1 and bay2 are only possibilities.

2. You make the interconnections so there is a route always awailable. modules have interconnection to module next to it and then you have created link from bay1 to bay3 and from bay2 to bay4. Any of the modules can fail and the three will have connection to each other

3. Any combination of 1Gb and 10Gb cables can be used to interconnect the Virtual Connect Modules
Remember to give Kudos to answers! (click the KUDOS star)

You can find me from Twitter @JKytsi
The Brit
Honored Contributor

Re: VCM fail over question.

Hi Jarkko,
Just to clarify Q2. Consider a blade which is talking on a specific vLan through one of the integrated NICs, and therefore is hardwired to IC 1 or IC2. To get to the external network however, it can only use an uplink which is available on IC3 and IC4. The path from the server to the network therefore looks something like;

Server -> IC1 (or IC2) -> IC3 (or IC4) via the Stacking Link -> Network.

In this situation, is the Server cut off when IC1 and IC2 modules die?

My feeling is that the answer is yes.

Clearly, this is not physically the same as removing the modules in IC1 and IC2 (which would remove the physical connections between the server NICs and the uplink). In this case the modules are still in place and so the question is really whether the physical link is still available even though the module may have failed.


Regarding Q3. Can you point me to any documentation which describes how to set up bundles of 1G uplink ports as a stacking link.

Thanks for your help and opinions.

Dave.
Adrian Clint
Honored Contributor

Re: VCM fail over question.

1. VMMC runs on VC1 or VC2... thats it.

2. You should design stacking cables in a loop. Which you have. So if a VC module fails then traffic should go flip to clockwise/anticlockwise the opposite to the route it took before. Obviously it cant contact the NIC on the downlinks of the failed VC.

3. Yes you can use multiple 1GB links as a wider link. It happens automatically aparently according to the VC docs. What you could do is have say 3 going left/right 3 going up/down and 2 diagonal. Then you have multiple stacking paths. You need to look at your traffic between NICs to plan the widest paths for the most suitable NIC use in your environment.
JKytsi
Honored Contributor

Re: VCM fail over question.

3. It says so in VC module quick specs =)
Remember to give Kudos to answers! (click the KUDOS star)

You can find me from Twitter @JKytsi
The Brit
Honored Contributor

Re: VCM fail over question.

Thanks for your help guys.