Aug 28

VMware vSphere 5.1 is new and shiny and doesn’t use VRAM!

Hardware version 9 and the naming of these versions is getting fixed up.
64 way SMP
1 TB VM memory
The vCenter Infrastructure Client is going to be end of life. The new web client is what you’ll need to get used to. Remember the web client from VMware 2 lol.
New tools no longer requires tools reboot.
vMotion without shared storage
Enhanced distributed switch
Storage DRS and profile driven storage integration with VCD.
Enhancements to Auto Deploy
DVS now supports LACP, Configuration backup and restore.
2x VM size shows VMware is serious about running all workloads in VMware.
VDP – VMware Data Protection, built on Avamar. Has some restrictions like ten virtual appliances and each appliance can protect about 100 VM’s. Replaces vSphere Data Recovery.
vSphere Replication uses virtual machine level replication. Included with vSphere. Does integrate with Site Recovery Manager and does support auto fail back.
vShield Endpoint lets you offload av scanning, uses 3rd party appliance.
New vCenter web client supports tags, pause and resume tasks, support for IE, Firefox, and Chrome. No safari.
Single sign on support for AD, Open LDAP, NIS, and local OS users.
Windows 8 server and desktop support
Express and Full install models for new vCenter server
VRAM is gone

My thoughts…..
I think auto deploy is great. I’m done patching hosts. I’m done troubleshooting them. Take a lesson from the pc guys who are just re-imaging your desktop when it breaks. Do the same thing on your hypervisors.

I think vCenter Orchestrator is awesome. I have said for years it will end up being the glue that connects all of the VMware products together.

Aug 28

Stretching a vSphere HA cluster can be tricky business. This is a task I have been working on for about 3 years. Latency is really the key requirement, lower the better of course. ESXi management has a max latency of 10ms. This is with enterprise plus, without it you are down to 5 ms. Your synchronous storage replication link must be 5 ms or less.

There are 2 different architectures. The first is really active/passive storage. The second is active/active storage. Some other considerations are HA, DRS, and Storage DRS are all not site aware. Storage is not the only thing to pay attention to. Network considerations are a huge list. Routing (layer 3) and switching (layer 2) are some big issues overcome.

Increase your datastore heart beats to 4 from the default of 2 so that in the event of a site failure you still have 2.

In version 5 some improvements have been made for HA split brain. The master is responsible for machine restarts. Master failures now cause an election of a new master in about 15 seconds.

A slide is displayed that is work getting with some recommendations and explanations for isolation response.

Even after you build it, maintain the configuration. Keep up with your restart priorities and affinity groups. Keep in mind HA doesn’t honor vApp restart order.

Automation! Automate virtual machine deployments and validate vm placement. vCenter Orchestrator is a good, use your tool of choice otherwise.

Test worst cases and your fears.
Some good VMware KB articles to review are
HP 2020097
EMC 2007545
NetApp 2031038

3 key failures for storage are all paths down, permanent device loss, and split brain. Each means different things and have different recoveries.

No surprise that if HA fails to restart your VM, it will move on to the net VM. Make sure you are using some monitoring tools.

This session is one to review at home as well. The slides have many scenarios that explain what will happen and has some whys.

An increasing trend is to utilize a stretched cluster in a single data center for things like a storage failure.

Aug 27

Josh Atwell with Cisco is the presenter.

Learn the instruments.
Learn the music
Build a band
Be a rock star!!

End of session…….
Read the rest of this entry »

« Previous Entries