Archives for author Christian Mohn

Yes, this is YAEo­tYP, so if you’ve already read tons of them I apologize.

2011 — My per­sonal view

2011 has been a steam­roller of a year.The vSoup Vir­tu­al­iza­tion Pod­cast got aired the first time, and we’ve recorded and pub­lished 19 full episodes in the inau­gural year. I was awarded the vEx­pert title for the first time, and even got invited to Tech Field Day #6 in Boston.

In addi­tion to this, I wrote a white paper for Veeam, was included in the Server Vir­tu­al­iza­tion Advi­sory Board, joined Rick Vanover for a Veeam Com­mu­nity Pod­cast, and appeared in two video interviews.

One with Mike Lav­er­ick about the #VMTNSub­scrip­tion­Move­ment and one where Eric Sloof ambushed me with a cam­era while vis­it­ing Bergen.

Lots of exit­ing projects were started in 2011, includ­ing my Pow­er­CLI based automa­tion project for ves­sel instal­la­tions, migrat­ing from stand­alone ESX hosts to blade servers (HP c7000 + Vir­tual Connect/Flex10) in addi­tion to the nor­mal day to day oper­a­tions and after 8 years at Seatrans AS, I handed in my papers, mov­ing on to a new role for EDB ErgoGroup.

2011 vNinja.net Statistics

2011 was the first year this site existed, so I can’t really com­pare the traf­fic it has received with 2010, but based on the few months it existed in 2010 the traf­fic increase has been enormous.

2011 Facts


And that’s it for 2011. Per­son­ally 2012 looks even more promis­ing, and hope­fully my expo­sure to more diverse envi­ron­ments should be reflected back on the site as I’m cer­tain it will spur more posts and more inter­est­ing dis­cus­sions. See you in 2012, I think we’re in for a cracker.

One of the last projects I’ve been involved with at Seatrans, is to auto­mate the instal­la­tion and con­fig­u­ra­tion of vSphere ESXi 5 hosts for deploy­ment on ves­sels. I’ve talked a bit about this before, both on vSoup and in Set­ting Up Auto­mated ESXi Deploy­ments where I out­lined my PXE and Pow­er­CLI based instal­la­tion and con­fig­u­ra­tion scheme. Not much has changed since then, except updat­ing the PXE server to offer ESXi 5, instead of ESXi 4 and a lot of work has been put into the script­ing, includ­ing a front-end GUI for the Pow­er­CLI script itself. The end “prod­uct” is now in place for mass deploy­ments for inter­nal use.

The fol­low­ing video shows how the PXE based instal­la­tion works, as well as a run through the now GUI based con­fig­u­ra­tion tool aptly called Seatrans Hyper­vi­sor Instal­la­tion Tool.

The video jumps a bit between two VMs, one run­ning Win­dows Server 2008 R2, that runs the DHCP/PXE ser­vices and the Pow­er­CLI script, and one that gets ESXi installed and configured:

This goes to show that you can cre­ate your own, spe­cial­ized and portable deploy­ment solu­tion with­out requir­ing elab­o­rate net­work con­fig­u­ra­tions or recon­fig­ur­ing of exist­ing infrastructure.

Note: I will not be pro­vid­ing down­load­able ver­sions of the final script at this time. The rea­son for this is quite sim­ple, it’s very spe­cific and tai­lored for a non-generic envi­ron­ment. If I can man­age to find the time, I’ll post a generic ver­sion later but in order for any­one else to uti­lize the Pow­er­CLI scripts I’ve cre­ated, a lot of work is required.

While using Thi­nApp to cre­ate a stand­alone ver­sion of Tweet­Deck 0.38.2, since the newly announced 1.0 ver­sion looks, acts and feels like a 0.1 ver­sion, I posed the fol­low­ing ques­tion on Twitter:

“Hrm, what other apps should i #Thi­nApp while I’m at it?”.

Kevin Kelling imme­di­ately responded with “Doom”.

Nat­u­rally, I decided to give it a go. A quick down­load of ZDoom later, and a quick run through the Thi­nApp Setup Cap­ture later, the fol­low­ing was born (view in full screen for bet­ter viewing):

Thus, DaaS (Doom as a Ser­vice) is born as a concept.

As the title says, it’s been one of my more “pub­lic” weeks ever. Besides my “nor­mal” vSoup engage­ment, this week I’ve also been involved with Mike Laverick’s VMTN Sub­scrip­tion Move­ment Mini­wags to voice some of my views about the #VMTNSub­scrip­tion­Move­ment.

Fair warn­ing: This is video, and please to remem­ber that dur­ing record­ing Movem­ber was near­ing its final phase.
VMTN Sub­scrip­tion Move­ment Mini­wags – Chris­t­ian Mohn

Sec­ondly, I was a guest on the Veeam Com­mu­nity Pod­cast Episode 45 – vSphere 5 Stor­age Pot­pourri.

Third, and last, Search­ServerVir­tu­al­iza­tion posted VMware vSphere Stor­age Appli­ance: Devil’s in the details which also includes some com­men­tary from yours truly regard­ing the VSA.

A lit­tle while ago I fit­ted a small 64GB SSD disk to my HP MicroServer just to have a quick look at the new vSphere 5 fea­ture Swap to Host Cache, where vSphere 5 reclaims mem­ory by stor­ing the swapped out pages in the host cache on a solid-state drive. Nat­u­rally, this is a lot faster than swap­ping to non-SSD stor­age, but you will still see a per­for­mance hit when this hap­pens. For more details on Swap to Host Cache, have a look at Swap to host cache aka swap to SSD? by Dun­can Epping.

Now, in my minis­cule home lab set­ting it’s some­what hard to get some real tan­gi­ble per­for­mance met­rics, so my lit­tle exper­i­ment is non-scientific and only meant to illus­trate how swap to host cache gone wild would look in a real world environment.

After installing the SSD drive, and con­fig­ur­ing Swap to Host Cache, I cre­ated two VMs inge­niously called host­cacheA and host­cacheB. Both were con­fig­ured with 14GB of mem­ory, which should nicely over­load my host that has a whop­ping 8GB of mem­ory in total.

Now, with mem­ory fea­tures like bal­loon­ing, trans­par­ent page shar­ing, and mem­ory com­pres­sion I needed to make sure that the actual mem­ory was used, and in addi­tion it had to con­tain dif­fer­ent datasets to make sure that the host cache actu­ally kicked in.

To make sure of this, I down­loaded the lat­est ISO ver­sion of Memtest86+ and con­nected it to the empty VMs.

When start­ing the VMs, they imme­di­ately started test­ing their avail­able mem­ory and sure enough, they started eat­ing into the host cache.

As you can see from the screen­shot below, the longer the memtest ran the more host cache was uti­lized.
Bonus points for fig­ur­ing out when the test VMs were shut down…

So there it is, per­for­mance graphs show­ing that the host cache is indeed kick­ing in and get­ting a run for it’s money. Since this was a non-scientific exper­i­ment, I don’t have any real per­for­mance coun­ters or met­rics to base any sort of con­clu­sion on. All I was after was to see if it came alive, and clearly it did.

VMware has announced Hori­zon Appli­ca­tion Man­ager 1.2, and together with the new Thi­nApp 4.7 release it promises “end users access to Win­dows, SaaS and enter­prise web appli­ca­tions across dif­fer­ent devices while retain­ing con­trol and vis­i­bil­ity via policy-driven man­age­ment”.

VMware Hori­zon Appli­ca­tion Man­ager now man­ages your Thi­nApp appli­ca­tions mak­ing it eas­ier and faster to pro­vide vir­tu­al­ized Win­dows appli­ca­tions to end users. From Hori­zon Admin­is­tra­tion, you can deploy Thi­nApp pack­ages, enti­tle users and groups, track user licenses, and man­age appli­ca­tion updates.

The cou­pling of the Hori­zon Appli­ca­tion Man­ager with Thi­nApp is a great idea, and when I saw today’s announce­ment I got pretty excited. The pos­si­bil­ity to have your own inter­nal appli­ca­tion por­tal pro­vid­ing your end users with self-service installs of vir­tu­al­ized appli­ca­tions is great news and could poten­tially be really use­ful in a great num­ber of organizations.

Sadly my ini­tial excite­ment quickly faded when I real­ized that for now Hori­zon Appli­ca­tion Man­ager is a hosted ser­vice, that requires an on premise con­nec­tor in your infra­struc­ture that sends over a lim­ited set of Active Direc­tory data to enable it to check user account or group access to the appli­ca­tions it offers. The con­nec­tor pro­vides sin­gle sign-on (Ker­beros) func­tion­al­ity for users already authen­ti­cated in your Active Direc­tory and authen­ti­cates the user to the Hori­zon ser­vice using SAML, so the hosted ser­vice never has the AD pass­word. The hosted ser­vice does still needs some infor­ma­tion like samac­count­name, first name, last name, email and a Glob­alUID.

For more details, have a look at Under­stand­ing VMware Hori­zon Appli­ca­tion Man­ager by Eric Sloof.

This also means that users who run a vir­tu­al­ized appli­ca­tion pro­vi­sioned by Hori­zon Appli­ca­tion Man­ager an active inter­net con­nec­tion is required, even if the vir­tu­al­ized appli­ca­tion pack­ages are stored on a local (to the user) file share. Sub­se­quent appli­ca­tion launches does not require an active con­nec­tion, as the appli­ca­tions are copied to the local sys­tem on the ini­tial run. The Hori­zon agents retrieves a lease for the appli­ca­tion, from the Hori­zon ser­vice, for an admin­is­tra­tor con­fig­urable num­ber of days (30 days default) and the end-user can run the appli­ca­tion, with­out con­nect­ing to the Hori­zon ser­vice, until the lease expires or is renewed.

For many orga­ni­za­tions, includ­ing mine, this poses a real prob­lem. “Hand­ing over” Active Direc­tory data to a hosted ser­vice is not some­thing I would want in my envi­ron­ment, espe­cially when our use case would be to pro­vide end users with a self-service appli­ca­tion por­tal for local appli­ca­tions. Other orga­ni­za­tions might look at that dif­fer­ently though, and this might not be a con­cern for all customers.

I under­stand that Hori­zon Appli­ca­tion Man­ager was ini­tially cre­ated for SaaS sce­nar­ios where a hosted authen­ti­ca­tion por­tal makes sense. I also under­stand that this is the first ver­sion that pro­vides inte­gra­tion with Thi­nApp, and this is very much a prod­uct still in devel­op­ment and refinement.

For now, Hori­zon Appli­ca­tion Man­ager does not pro­vide the use case that I was look­ing for but thank­fully Ben Good­man, Lead Evan­ge­list for VMware Hori­zon, has taken the time to address my call for an on-premise ver­sion of Hori­zon Appli­ca­tion Manager:

I under­stand your appre­hen­sion. Hori­zon was built on top of tech­nol­ogy orig­i­nally designed exclu­sively to be a Single-Sign on ser­vice to SaaS appli­ca­tions. We are in the process of expand­ing that tech­nol­ogy to become a true enter­prise ser­vice. This is hap­pen­ing in two ways, the first is by adding appli­ca­tion sup­port beyond SaaS. The first step was Win­dows sup­port via Thi­nApp and we are look­ing at other appli­ca­tion plat­forms to fol­low. The sec­ond is eval­u­at­ing options for mov­ing some or all of prod­uct on-prem. Both of these steps are the pri­mary focus of the devel­op­ment team over the next 12–18 months and we are really excited about where we are tak­ing Horizon.

This is great news, an on-premise ver­sion that pro­vides exactly what I’m look­ing for seems to be in the pipeline and on VMware’s roadmap for Hori­zon Appli­ca­tion Man­ager. I just wish I had it now, it would have been per­fect for a project I’m work­ing on at the moment that I hope to wrap up by the end of the year.

Oh well, there is always next year and the next project!

Mike Lav­er­ick has started some­thing of a peti­tion to bring back the VMTN Sub­scrip­tion option, and I could not agree more!

The VMTN Sub­scrip­tion was a way for inter­ested par­ties to pay for a years sub­scrip­tion to VMware prod­ucts, akin to the Microsoft Tech­net sub­scrip­tion pro­gram. It’s not intended for pro­duc­tion use, but as a means to get hold of prod­ucts for lab work, test­ing and development.

I don’t under­stand why VMware pulled the plug on that option back in 2007, but I do under­stand why it’s time to bring it back to life. As is the case with Mike, as a vEx­pert I can prob­a­bly get hold of all the bits and pieces on my own, but not every­one has the same oppor­tu­ni­ties and I’m sure that’s actu­ally sti­fling com­mu­nity knowledge.

The VMware com­mu­nity is filled with great resources, avail­able and shared between it’s mem­bers, all we want is for VMware to enable the com­mu­nity to grow even more by facil­i­tat­ing home labs, test envi­ron­ments and explo­ration of their prod­ucts. That has to be in VMware’s own inter­est too.

It’s not like we’re look­ing for a free lunch here, but what we’re look­ing for is some­thing between the 60 days trial ver­sions and the full pro­duc­tion licensed prod­ucts. After all, we don’t like rebuild­ing our labs every 60 days do we?

The announce­ment of the VMware Labs going pub­lic in 2012 is a step in the right direc­tion, a rein­state­ment of the VMTN Sub­scrip­tion would be another big step. Come on VMware, I know you have it in you!

If you want to add your own voice to the dis­cus­sion, have a look at this VMware Com­mu­ni­ties dis­cus­sion thread, it’s already got some trac­tion and the more atten­tion it gets the better.

Juan Manuel Rey’s post Mon­i­tor ESX 4.x to ESXi 5.0 migra­tion process show how you can watch the progress of an ESX 4 to ESXi 5 upgrade pro­ce­dure, by look­ing at the live logs.

While this is very use­ful, and in many cases a real learn­ing expe­ri­ence, it got me think­ing that these logs should be avail­able remotely as well. Since ESXi sup­ports, and actively encour­ages, the use of an exter­nal Sys­log ser­vice for log file safe­keep­ing and mon­i­tor­ing, shouldn’t the instal­la­tion logs for ESXi also be logged exter­nally if configured?

Think­ing that I couldn’t be the first per­son that thought of this, I looked through the script­ing sec­tion of vSphere Instal­la­tion and Setup — vSphere 5.0 guide I was very sur­prised to see that there is no option to con­fig­ure sys­log­ging until after the instal­la­tion is fin­ished and the host con­fig­u­ra­tion script(s) runs (ks.cfg).

By using a ks.cfg script you can auto­mat­i­cally con­fig­ure sys­log set­tings, but since that hap­pens after the instal­la­tion is done, and the host is poten­tially rebooted, the instal­la­tion logs are lost (ESXi logs are not per­sis­tent by default) unless you run some­thing that copies them over to another loca­tion before the reboot happens.

Of course, when I asked on Twit­ter the God­fa­ther of Ghetto, William Lam responded that you could always cre­ate a python script that runs after the install, before a reboot, that uploads the logs to a sys­log server. While this is all fine and dandy, I would still like to have the pos­si­bil­ity to con­fig­ure a sys­log server dur­ing the instal­la­tion, and have the instal­la­tion pro­ce­dure fling all it’s log good­ness at the sys­log server while the instal­la­tion runs.

With new fea­tures like Autode­ploy being uti­lized, hav­ing these logs auto­mat­i­cally gath­ered (with­out ghetto hack­ing the instal­la­tion) in a cen­tral loca­tion sounds like a really good idea to me?

Surely, I’m not the only one? Is this some­thing that has enough brains behind it to actu­ally war­rant a proper fea­ture request being filed with VMware?

Every once in a while an oppor­tu­nity presents itself that is just too good to pass up and after 8 years at Seatrans AS I’ve decided to move on and and accept a posi­tion as a Senior Con­sul­tant in the Infra­struc­ture Con­sult­ing divi­sion of EDB ErgoGroup.

Seatrans has been a fan­tas­tic employer, and with­out the back­ing and sup­port I’ve had over the years I would not be in a posi­tion where this change would be pos­si­ble. It was with mixed emo­tions I handed in my notice, but I’m 100% cer­tain that this is the right move, at the right time, for me personally.

As a con­sul­tant my main focus will still be vir­tu­al­iza­tion in gen­eral, and VMware solu­tions in par­tic­u­lar. The upside to this is that I will def­i­nitely be able to work even more with the tech­nol­ogy clos­est to my heart, and with a larger team that has a sim­i­lar pas­sion for tech­nol­ogy. Mov­ing back to the con­sul­tant side of the table should be an inter­est­ing chal­lenge, and one that I’m really look­ing for­ward to.

The blog will stay the same, per­haps this change might even bring more con­tent in as I’m going to be exposed to a lot of dif­fer­ent infra­struc­tures and challenges.

I’ve also set a cou­ple of pretty hefty per­sonal goals for 2012, but I’ll keep those to myself until I see how every­thing pans out.

2012? Bring your A game, because I’m awake, strong and ready!

if ((Get-date) -gt (Get-date 2012-01-01))
    {Get-VM h0bbel |
      Move-VM -Datastore newEmployer -RunAsync
     }

More details later.