Archive for September, 2009

I did a lot of various different things this week. It was one of those weeks where I did loads and loads of different things – and got to the end of the week – and wonder where the hell it had gone – and what I’d actually achieved… With that said there was two stand-out events this week. I did two remote sessions with Alex Tanner of EMC. Alex appears to be Chad Sakac’s right-hand man in the UK. A stand-up fellow who really knows his stuff. You know the kind of guy who tells you stuff – and you have sit back think – wow, what are the implications of all this stuff.

What prompted this delve into EMC is that I’m due to User Group presentation at the beginning of October – and I’ve picked the theme of Storage Integration – specifically looking how various vendors have adopted the new vStorage API to add additional functionality to vCenter. In reality vStorage is much bigger than just giving you a fancy icon in vCenter or a some extra menus – in the longer run it’s going to allow for much more intelligent interaction between the ESX host and your array. I’ve picked on EMC and NetApp as example of this integration, it also helps I’ve got their kit racked up in my co-location which means I can play to my hearts content – plus I’ve good contacts at both companies – to ask them additional questions.

I’m hoping I will have time to spend a couple of hours with the NetApp guy I know – to do a similar crash course in additional addons. For the moment I’ve only played with NetApps RCU but I know there is more to be seen because I’ve got the software on my shared drive waiting for me to find the time to install it…

EMC Storage Views: (FREE)
First up is storage views – it’s a vCenter4.0/vSphere4 only plug-in which gives you much better insight to your array from vSphere4 Client – it also gives you additional info about your VM. It’s a total doddle to install. You install the Solution Enabler from EMC to your client machine, and then crank-up the Storage Views install. Your talking 5min job. After this you get a nice little EMC logo in vSphere Client like so:

When you click at this icon, you tell it of the Clarrion systems you have by IP address of SPA/B and the username/password to authenticate:

The your done. This will add an EMC Storage Tab to the Host & Cluster view in the vSphere Client.  So in the screen grab below. I’ve select the “isos” LUN on one of my CX-3 Clarrions. Notice how EMC Storage Views – allows me to map the VMFS datastore the LUN name and storage group as it is known in Navisphere. That’s really helpful to me – as sometimes I get little confused about the mapping of the LUN to VMware, and vice-versa.

You get a similar view on a properties of VM too

EMC PowerPath:

Something I’ve blogged about already is PowerPath for VMware. If you looking for a step-by-step guide on setting it up, installing and licensing it. Look here. It’s dead easy to do as well:

http://www.rtfm-ed.co.uk/?p=1360

EMC MirrorView for VMware:

Another plug-in comes with the new Site Recovery Adapter from EMC which is compatiable with SRM 4.0 which is currently in beta. The EMC MirrorView for VMware will carry out a knowledge consistent check on your set-up – including checking the part of the VM aren’t orphaned. It can also automate the fail-back process.

EMC Fast Clone:

This was previewed at this years VMworld. It’s essentially a replacement for the Vi3.5 VDI Deployment Tool. The new “Fast Clone” utility looks a bit more intergrated than the previous utility.

CX4 and NaviSphere:

I was also given a sneak preview of the new version of Navisphere that runs on the CX4. From Navisphere you get a view of the virtual machine from the storage management tools. See it like the other end of the telescope of EMC Storage Views. So with Storage Views you see your array, and from the new version Navisphere – you can see the ESX hosts and VMs. So whether your a VMware guy, a Storage guy or both things are much more joined up now. So you a small (H) icon indicates an ESX host, and you can expand it to show the VMs on that ESX host:

The new version of Navisphere, has search option to locate a VM, and then you can crack it open to examine the virtual disk set-up

There’s also a capacity to run reports specifically angled at VMs:

EMC Replication Manager:

I covered all this stuff with Alex on Wednesday, and then today we spent some time setting up EMC Replication Manager. Basically, its a windows service that serves as proxy between your ESX hosts and your arrays. Although its much bigger than just that because there application specific agents for things like SQL and Exchange. Alex took me through the set-up which didn’t take too long. We set-up through Replication Manager a snapshot of my “VIDEOS” VMFS volume.

Once the snapshot process is in place you can mount the snapshot. Using vSphere’s new capability to re-signature volumes via the GUI when you run the Add Storage wizard like so:

After doing the re-signature, the ESX hosts are automatically re-scan to display the original volume and the snapshot volume. After that it’s more or less a copy and paste process to get the data out of the snap volume into the original volume:

As the post says. It’s available…

Well, there’s seems to be a little mission creep involved in the cloud circles. It seems that VMware’s public stated aim is this. VMware is vendor of products, which can be used to build either an internal or external cloud – what it hopes and some degree has already successfully done – is to convince some 1,000 providers to adopt vSphere4 as the platform for delivering the cloud via its vCloud Express initiative.

At this months VMworld, during the vExpert session I asked whether VMware had any intention of producing a cloud service on its own. The reason being is that I’ve rumours and reports of rumours that VMware has bought more datacenter capacity than it actually needs to run its own infrastructure. The person I asked was VMware’s CTO Steve Herrod. He flatly denied that VMware had such a plan. And I believe him – mainly because you would encourage 1,000 companies to adopt your platform to deliver the Infrastructure-as-a-service (IAAS) to then compete against them. Would you? That would be daft wouldn’t it? Not only would have annoyed a lot of partners, you would overnight turn them into competitors at a stroke.

With that said, it does seem odd that if this was VMware’s publically stated goal – that they would acquire such a significant stake in one of those vCloud Express providers - Terremark. Where does that acquisition fit into the mission statement?

http://www.infoworld.com/d/virtualization/vmware-buy-20-million-stake-in-terremark-835

I thought I would go digging about on vmware.com website to look for announcements about the buy into terremark and found this on news releases:

http://www.vmware.com/company/news/releases/vam-vsphere-launch.html

“VMware Enables Users to Easily Test-Drive Cloud Computing through the VMware Virtual Appliance Marketplace (VAM) and VMware vCloud™ Service Provider Free Trials”

Perhaps the reasoning behind the acquisition – is that VMware could learn about the realities of deploying its products with an organisation like Terremark. It could also push the agenda of cloud computing on more successfully by buying into one of the many 1,000′s of providers – using them to showcase bleeding edge technologies and ideas. After all they only acquired $20m stake which sounds a lot of money but is nominal amount in real terms – it represent 5% stake when VMware bought some 4 million shares at $5/share…

What prompted this blog post was this blog post on cloudscaling.com

http://cloudscaling.com/blog/cloud-computing/vmware-vs-amazon-round-one-fight

It’s an interesting read. Essentially, the author thinks that standards will trump Amazon’s market share – that the openness VMware’s vCloud API (it even, shock horror, allows for hypervisors other than ESX to be used) will win out against the currently closed model of Amazon EC2. Of course, Amazon could open up its standards to stop the “Hotel California” affect. But if it doesn’t do that quickly it will find that other cloud providers adopt the VMware standard before them. Then the predict move by Amazon to Xen will be regarded as too late. It seems like that big business which trust and already have VMware will first run private internal clouds, before they even consider public external clouds – and the move to a public external cloud will seem less “risk” if the technology behind it has already be tried, tested, broken and fixed – inhouse, before its is outhouse.

Of course this outhouse has other connotations entirely…

http://en.wikipedia.org/wiki/Outhouse

Not sure if this is new, but its new to me. You can sign up for email notifications for patch/maintenance alerts here:

https://www.vmware.com/mysupport/subscription/

The odd thing is the decidely lack of options for ESX4/vCenter4

VMware has announced a new blog – just on vSphere4. To celebrate this fact and to seed the new blog with community content they have launched a contest. Every two weeks there will be topic – and people will be asked to blog about – the best post wins a coverted listing on the new vSphere blog, and win a $100 AMEX card…

http://blogs.vmware.com/vsphere/2009/09/kicking-it-off-vmware-vsphere-blogging-contest.html

The first topic is about fault tolerance. Ever the devils advocate, I’m sitting on a blog post called “What sucks about VMware Fault Tolerence”. Not sure if I will post it – as it might upset people I know at VMware too much. Generally, I really like FT, but its not without a truckloads of challenges and limitations. But I’m thinking the contest isn’t really designed to flag up negatives, but positives…

You might recall I didn’t actually make it to the second days keynote. I was actually lying on my bed in the Marriot recovering from the previous nights exercisions. I took to monitoring twitter and assembling a blog post even though I wasn’t actually at the keynote. How’s that for virtual blogging.

Anyway, I heard the was some news about DRS for Storage in the keynote – and I was intrigued enough to actually watch the session on VMworld 2009 where the sessions have recently become available.

http://www.vmworld.com/community/sessions/2009/

My plan is to watch one of these videos each day and then blog about them. Given I only made it two sessions (and one I had to be at because I was speaking) I thought I should pay some penance for my sins.

Anyway, if keynote’s aren’t you bag – your welcome to skip this. As ever you have to work pretty hard to get something concrete from a keynote. In the past however, the keynotes have been used to showboat bleeding edge technologies – the best for being the demo of Continious Availability (AKA Fault Tolerence) made by the founder of VMware in 2007…

(more…)

The sessions from VMworld from a couple of weeks ago are now available online – for free for those who attended the event physically, and for subscription fee if you didn’t. To subscribe toddle on down to:

http://www.vmworld.com/community/subscription

And get ready to part with $699 for years worth of access…

If you attended – just go here and login:

http://www.vmworld.com/community/sessions/

I only attended one breakout session, spoke at another – and only managed it to one key note. The rest of the time I was pressing the flesh and doing meetings (Ed, lottering around the solutions exchange without intent?). My intention (remember the road to hell is paved with those) is to go through my ones from my session builder – producing an executive summary of each one – with the main take aways…

Blogger/Tweeter Danile Eason of www.vmlover.com has written a free to download (you must register first) overview guide to Cisco UCS… It’s hosted on the popular xtravirt.com website here:

http://xtravirt.com/xd10120

Here’s da blurb:

“UCS (Unified Computing System) is Cisco’s first entry into the x86 server market. This paper provides insight into both the Cisco Data Centre 3.0 strategy, and UCS, which is a key enabler to DC 3.0.  It discusses products within the UCS range at a component level, such as blade and rack mount server offerings, and lastly highlights some of the larger benefits that can be gained through the pairing of x86 virtualisation and UCS technology.”

[Er, actually spell checked this (for once) so removed the very worse of my typos/spelling...)

I was training last week - VMware View product. To be truthful it’s not product I’m especially enamoured off. I find its admin tool crude and limiting especially… users don’t seem to mind it however, but then they don’t really give a crap what the broker is, so long as they get their desktop and email. Anyway, I’m expecting/hoping for good things in the next release - not that I would know anything about that - I can’t get on the beta programme despite having more v’s after my name than is really seemly (vExpert, VCP, VCI, vImportant).

Anyway, grumbling aside - one of my student pitched a usage case to me and asked me if I thought VMware View could deliver. In the first instance you would want to recommend “Lab Manager” for this guys usage case - but I know that a budget for yet-another-management-product-from-vmware would probably not be there - it was up to me to trying cobble something together. Here’s the usage case. He wants his developers to get a desktop - install stuff and generally **** around with the desktop - but then give them an easy way to reset the desktop back to clean state. At the moment they have vCenter access to use snapshots/revert to snapshot - but they aren’t keen on giving them access to vCenter… They want to do it via View. Of course, non-persistent pools could help in this case - they have a feature which allows you to delete (and create a new) a virtual desktop at logoff. But both myself and this guy could see that rather than disconnecting one day a developer would logout - to find out their work was destroyed. I guess a bit of redirection and hardening could force them to store their stuff outside of the desktop was a possibility - but my guy was trying to aviod bolting them down…

So I hit upon this idea:

• Create a persistent pool of desktops
• Let them power up and sysprep themselves
• Use PowerCLI to shut them all down
• Change the Disk Mode on the OS drive to be independent mode with non-persistent
• Set the Desktop Pool to “Ensure VM is always powered on” and “Allow users to reset desktop”

This means if the developer in the virtual desktop hard shutdown their desktop OR use the Reset Desktop option in the  View menu. Their desktop would be reset to its previous state

This works fine as a concept but falters in two places. If the developer did a restart of Windows with the NT Security dialog box – they View Agent didn’t appear to restart properly (I figured I could write a powershell script to handle that which would run within the developers virtual desktop – the downside of which leaves me into giving them access to vCenter – which if you recall my guy was trying to aviod…)

The second place it falters in the use of non-persistent mode on a virtual disk. I was reading through the help file on the Set-HardDisk (which supports changing the disk mode of a VM) cmdlet when I came across this statement:

“Specifies the disk persistence mode. The valid values are: Persistent, NonPersistent, IndependentPersistent, IndependentNonPersistent, and Undoable. This parameter is supported only when the disk type is set to “rawVirtual” or “flat”. Note that the ‘NonPersistent’ and ‘Undoable’ persistence policies are deprecated and will be discontinued in future releases. Their usage is not recommended because they do not work with snapshots and are not supported on ESX 3.5 and higher.”

This was taken from a machine running the new PowerCLI 4.0 which is compatible with vSphere4…

The operative word is “deprecated”. It’s a special word used by VMware which you rarely hear from other vendors. I put up there with the usage of the word “experimental”. Occasionally, VMware deprecated a feature – it’s a rather artful way of saying “You know that feature you liked, understood and wrote scripts and procedures for… well, my friend – were removing it.” Just like that. No consultation.

It’s not first time I’ve come across with VMware this – for example in ESX3.x.x the old ESX2.x.x command vmkpcidivy got deprecated because it was no longer relevant. Similarly, the functionality of esxcfg-vmhbadevs in ESX3.x.x got deprecated in ESX4.xx. favour of another utility that provides much better functionality.

You see I don’t have problem with “ deprecated” especially when a new method is better or the old method is simply irrelevant. Beside which I think we can all agree we don’t want software bloated out with old procedures/utilities/code which is no longer needed. That would make you like other ISVs which will remain nameless!

But what irks me about it is the arbitrary nature of it. You never quite know when a feature might get deprecated on you (such as the shink feature in VMware Tools). Often the deprecated process seems quite arbitrary, especially when it happens within a product release – if it happened between two major product releases that could be understandable/excusable. But when it happens within a product release occasionally, it feels a bit odd. Can you imagine the manufacturer of your car deprecated the wing mirrors of car after you have bought it?

Clearly, vendors do reserve the right to discontinue models, but they don’t tend to remove features during the lifetime of a model. So model 3.0.2 of a car has wing mirrors, but in the 3.5 upgrade which they forced you to move up to keep your service plan they deprecated it.

Anyway, back to my powershell. I decide to ignore the advice. I used:

get-vc virtualcenter4.vi4book.com –user administrator –password vmware

stop-vm -vm (get-vm -name devdesktop*) -confirm:$false

get-vm -name devdesktop* | Get-HardDisk | Set-HardDisk -Persistence IndepedentNonPersistent

In short you can configure non-persistent mode on virtual disk in the vSphere4 client, and you can do it with the PowerCLI. But you run the risk if use this feature of being deprecated at some unspecified point in the future…

Anyway, what to say to my customer? Set non-persistent using PowerCLI, and but beware of upgrades could stop these scripts/function working in the future?  Offering him snapshots powersHell scripts instead? (This is what I’ve done)  Sell him a product (lab manager) he doesn’t desire or have a budget for?

Your deprecatedly
Mike Laverick….

Well, this particular configuration has been outstanding for me for more than month – and I’ve finally cracked it (albeit with a little help from my friends… beatles referrence geddit…?) Borrowing from Carter Shaklin’s example of enabling FT Logging – I found a method of doing this:

You have to be little bit careful when you run this script. You don’t need this script with ESX “Classic” which uses the Service Console. For that you can use the -consoleNIC parameter to make sure the new VMHostNetworkAdapter is a vswif interface rather than a vmkernel interface

$vs3 = vSwitch3
$HAheartbeat = New-VirtualPortGroup -VirtualSwitch $vs3 -Name HA-Heartbeat
New-VMHostNetworkAdapter -PortGroup HA-Heartbeat -VirtualSwitch $vs3 -IP $HAheartbeatIP -SubnetMask 255.255.255.0 -ConsoleNic

To do the same for ESXi you have dig into the SDK using the data object virtualNicManager held with configManager to set the “Management Traffic” attribute.

$hView = Get-VMHost esx4.vi4book.com | Get-View -Property configManager
$nicManager = Get-View $hView.configManager.virtualNicManager
$nicManager.SelectVnicForNicType("management", "vmk2")

The only slightly annoying thing about this PowersHell is don’t think I would have found these parameters without the help of the PowerCLI product manager – I’m trying to work out how he found these values – it could wind-up being an internal engineering thing, than something in the public domain. Well, it is public now

I suspect the that SelectVnicForNicType companion QueryNetConfig was used to find out the variables on an existing configured ESXi host to find out the values for ”faultToleranceLogging” and “management”. I wasn’t so clever, I worked out “management” by guessing!

Update:

Anyway, later after posting this the Product Manager for the PowerCLI ping’d me an email with URL showing where the values of “management”, “vmotion” and “faultToleranceLogging” were held on the

http://www.vmware.com/support/developer/vc-sdk/visdk400pubs/ReferenceGuide/vim.host.VirtualNicManager.NicType.html