Comments for vNephos.com

Comment on NX-OS port-channel hashing by Andy Daniel

Andy Daniel — Sat, 28 Jan 2012 17:16:50 +0000

The command doesn’t alter the algorithm in any way. It simply shows you the interface that would be used if that traffic flow traversed the port-channel. It can be very useful when planning your IP address assignments.

Comment on NX-OS port-channel hashing by darul

darul — Sat, 28 Jan 2012 16:42:03 +0000

regarding the cmd that you mentioned, actually what the purpose of the command itself? is it altering the load balance algorithm or showing about the load balance for the port channel? sorry because im still new with this nexus. Hope u can brief me. TQ

Comment on Nexus 55xx & 50×0 vPC incompatibility by Technology Short Take #20 - blog.scottlowe.org - The weblog of an IT pro specializing in virtualization, storage, and servers

Fri, 20 Jan 2012 10:39:16 +0000

[...] looks like you can’t create a vPC (virtual port channel) between a Nexus 55xx and a Nexus 50×0 [...]

Comment on HP ProCurve cross-stack EtherChannel by Erki

Erki — Wed, 21 Dec 2011 14:00:48 +0000

At least E5406zl using K.15.07.0002 fw has the option to configure dt-trunk instead of dt-lacp if needed. Works great with ESXi Active/Active with ip hashing.

Comment on Another Nexus 1000v WFQ QoS Example by Martin

Martin — Mon, 19 Dec 2011 22:33:39 +0000

Missing a few steps of configuration there?

How do you use the match protocol along with the set cos in the same type of policy/class-map?

Nested policies?

Cheers!

Great post by the way – I also agree on the need to use the 1k to mark the traffic (COS) specifically (unlike the catalyst, which can convert DSCP automatically) to be used on the layer2 network!

Comment on HP ProCurve cross-stack EtherChannel by Andy Daniel

Andy Daniel — Sat, 20 Aug 2011 00:25:27 +0000

Steven, sorry for the late reply on this. It somehow got lost in the shuffle. I”m sure you’ve probably had to deploy your solution by now, so my apologies. What I’ve learned since this original post is that the distributed trunking mechanism does indeed rely on LACP. This means from a VMware perspective the only way to make it work would be to use the Cisco Nexus 1000v. Kind of ironic, isn’t it. You can still use the 6600 switches with ESXi, just not Active/Active with IP Hashing. The normal Active/Active setup, that effectively uses mac-pinning, will work just fine. You could also use the two NICs in an Active/Standby configuration, still get use from both, and more closely control which traffic uses which NIC. In this scenario, simply set one NIC to be Active for some portgroups (say Management & vMotion) and the other to be Active for others (like iSCSI). In this scenario you’d set the opposite as the Standby. Some in the industry refer to this configuration as “poor man’s load balancing,” and its done by simply overriding the Teaming settings per portgroup within the same vSwitch. It works very well.

Comment on HP ProCurve cross-stack EtherChannel by Steven

Steven — Mon, 16 May 2011 16:06:28 +0000

Hi there,

I’m wondering if there has been any update since February on this? We are about to deploy a new ZFS based SAN, and had originally planned on connecting our small 3 host ESX environment via two 2910al switches.

Upon discovering that the 2910al switches do NOT support distributed trunking, I have planned to replace them with 6600 series switches instead.

My main issue is that I plan to put iSCSI traffic on 10GbE interfaces – 2 from the SAN, one interface to each switch, then connect each host via 2 10GbE interfaces, one interface to each switch. I had hoped to have an Active/Active connection, but at this point would live with Active/Standby as we won’t exceed 10GbE anytime soon.

Now I’m not sure that I’m really gaining anything by moving to the 6600 series for my particular configuration.

My greatest concern is how the failover appears to be problematic with ProCurve + ESX environments where multiple switches are used. I need to know that if switch_1 goes down, the Guest VMDK files will seamlessly remain available to the hosts via switch_2.

What isn’t clear is if distributed trunking is required for this or not.

– Steven

Comment on Nexus 55xx & 50×0 vPC incompatibility by Charlie

Charlie — Sat, 23 Apr 2011 04:07:23 +0000

Very good to know! Wish I would have known this before starting a migration! Thanks for posting.

Comment on Cisco UCS C-Series, VMware & RAID by rob

rob — Tue, 15 Mar 2011 16:48:53 +0000

thanks for this, pointed me in the right direction! Cheers

Comment on HP ProCurve cross-stack EtherChannel by Andy Daniel

Andy Daniel — Thu, 10 Feb 2011 12:35:33 +0000

donkarnash, it sounds like you have issues outside of what we’re discussing here. Please feel free to email at adaniel at vnephos dot com, and I’ll be glad to help or get you in touch with someone who can.

Andy