I attended the VMworld 2012 session that covered the new features in vSphere 5.1 with regards to networking. Many features were rolled out to both VDS and the standard switch, and other features just had improved functionality.
First off, apparently it’s now VDS, not vDS. This announcement came hours after the announcement that VXLAN was being changed to vXLAN. Um…okay, I guess?
Anyways - The speaker pointed out at the beginning that a big change was that many of these features were being rolled out to both the standard and distributed switches.
I ran into an issue that presented itself two different ways, each at a different customer. I posted a while back about a customer that wanted to use only a single Nexus 5000, since that was all that was available. I wanted to bundle all four CNA ports on the Netapp storage array to the Netapp SAN. However, after I created this port channel and bound the virtual fibre channel (VFC) interface to it, the VFC interface would not come up.
I ran into an issue that presented itself two different ways, each at a different customer. I posted a while back about a customer that wanted to use only a single Nexus 5000, since that was all that was available. I wanted to bundle all four CNA ports on the Netapp storage array to the Netapp SAN. However, after I created this port channel and bound the virtual fibre channel (VFC) interface to it, the VFC interface would not come up.
I had the opportunity this week to ascertain the feasibility of automating the provisioning of a full Flexpod. For reference, this is considering a “vanilla” Flexpod build:
Pair of Nexus 5ks Pair of Cisco UCS Fabric Interconnects (with a few chassis) Netapp running ONTAP 7-Mode (I tested on FAS6070) Note that this also makes a few assumptions about the build.
FC via Nexus 5000, no MDS No existing vCenter integration or storage migration So - pretty much a green field Flexpod build, pretty close to the specs laid out in the design guide.
I had the opportunity this week to ascertain the feasibility of automating the provisioning of a full Flexpod. For reference, this is considering a “vanilla” Flexpod build:
Pair of Nexus 5ks Pair of Cisco UCS Fabric Interconnects (with a few chassis) Netapp running ONTAP 7-Mode (I tested on FAS6070) Note that this also makes a few assumptions about the build.
FC via Nexus 5000, no MDS No existing vCenter integration or storage migration So - pretty much a green field Flexpod build, pretty close to the specs laid out in the design guide.
This is a (justifiable) rant. You’ve been warned.
Solarwinds Orion NPM is an okay tool, but when it comes to managing anything other than Cisco switches and routers, it’s…..meh. It takes very little effort to get devices like that monitored to the fullest extent, but when it comes to something like a storage array, it seems like you really have to make tweaks until your fingers bleed to get the minimal monitoring functionality out of it.
The SAN I’m currently working with connects a pair of Netapp FAS3270 filers running ONTAP 8.0.2 7-Mode.
If you’re running VMware ESXi in your environment in front of a Fibre Channel SAN, path selection is discovered more or less in a first-come-first-served fashion.
I got this message on my Netapp filer:
FCP Partner Path Misconfigured: Host I/O access through a non-primary and non-optimal path was detected. Since the LUNs mounted by ESXi were residing on the A-side filer, the paths going through the B-side filer would just be sent over the partner link to the A-side, which is less efficient than going directly through A.
I came across this the other day and wanted to share. For some reason, Windows by default decided to enable the “audio enhancements” feature on my new Lenovo Thinkpad w520.
This caused my microphone to essentially be unusable - I was in several webex meetings and each time everyone said I was completely garbled and not even close to being able to understand me. After a little poking around, I found this:
Cisco announced recently a replacement program for their B440 blades (M1 and M2)
http://www.cisco.com/en/US/ts/fn/634/fn63430.html
Something I noticed on the front of each blade, where the model is shown, the new generation blades (replacements) have the black background with the silver text, shown as the top blade on the picture below:
However, the old blades still have the black text with no background. An interesting way to identify the newer generation of hardware.
If you’ve set up an ESXi host, you’ve likely seen this screen:
This allows you to configure which VLAN is used for management. But what does this really do? Time after time I run into very smart engineers that primarily work on virtualization and not as much on the physical networking side - and they miss a few of the networking fundamentals that those of us that were brought up in ROUTE/SWITCH know and love.
It’s been a while since I’ve done a review of anything on the site and since I’d rather continue to make use of the category than delete it, I decided to share a piece of software I recently discovered that’s helping make my life easier.
Remote Desktop Connection Manager is essentially just that - it manages remote desktop connections. However, it does it in a way that I find appealing and EASIER to use than the native client or other organizational methods out there.
For those that have worked with any type of blade server system, you know that boot from SAN is just about the coolest thing since sliced bread. Cisco UCS makes this even cooler by integrating with the service profile concept, allowing for stateless compute provisioning across the board.
I’ve done boot from SAN many times, but never with Windows. I’ve primarily used ESXi4.1 or ESXi5.0 stored on a Fibre Channel LUN, then the VMs are stored in either a FC or NFS datastore.
Snapmirror is a Netapp feature that’s pretty commonly used to copy data from one system to another. You can copy volumes, or qtrees. It’s also very useful in Disaster Recovery plans, since volumes can be incrementally backed up to an offsite location.
I have a VMWare vCenter instance running on Cisco UCS that utilizes a Fibre Channel LUN to store VM templates. It’s pretty large, since it holds templates for a variety of operating systems.
I’m currently working with a relatively large Cisco UCS installation. Initially, the system was installed and brought up to relatively recent levels of firmware, but a mismatch in the way that the firmware packages were set up in various sub-organizations on some of the UCS systems caused some of the blades to retain the old version of firmware on the M81KR adapters and the CIMC controllers.
Due to the scope of the installation, I wanted to ensure that the blades were able to continue operating while I made my changes.
I recently observed some strange behavior with Cisco UCS Manager. When I visited the web page that allows me to download the .jnlp file that launches UCSM, it came up just fine. But when I clicked on “Launch UCS Manager” to actually launch this applet, the splash screen showed briefly, but disappeared after a few seconds, never to be seen again.
Eventually, you might also see some java error messages that say something like
I would like to share some tips regarding gadgets that I believe every Datacenter Network Engineer should have with them. There are several, but I want to bring up my top two.ß
Travel Router I am often in situations where it is either difficult or impossible to manage Nexus switches and/or UCS remotely. Pick your reasons - sometimes the management network doesn’t exist (yet) or there are heavy security measures in place that restrict wired management, whatever.
Static routes can be handy in some situations where you want to do some quick and (sometimes) easy routing to get the job done, whether replacing the job that a routing protocol would perform, or redistributing the static route into that protocol.
The best way to do this would be to identify the remote subnet being routed to, and specify a next-hop IP address to send traffic to so that it can be reached.
I ran into a configuration recently where I had a Netapp storage array with the UTA cards installed, so there two CNA ports on each filer for a total of 4 ports. However, instead of a dual-switch design, there was only a single Nexus 5000, and therefore, no vPC configuration. I needed to achieve some level of redundancy on an interface level, but ran into some problems which I’ll discuss.
Port-Channels, are a way of aggregating physical links together so that you can load balance traffic over each link to increase bandwidth, and create more redundancy. You might commonly see this configured between two switches, as shown below:
Each link works together to form a logical, loop-free interface. These are relatively commonplace, and in this scenario highly useful because it prohibits spanning tree from blocking one of these ports, allowing the switch to utilize each link.
It’s interesting to me to see the differences in infrastructure products as it pertains to out of the box, or default configuration. Take for instance, the relationship between a firewall and a switch. Your average firewall is configured “closed”, meaning that if you want to allow anything, you have to explicitly allow that certain type of traffic. If you do not, it is not allowed. A switch, on the other hand, is configured to be functional above all, out of the box.
Keeping It Classless