Managing Pods in a Horizon View Cloud Pod Environment using PowerCLI & Api’s

After Pod Federations and Sites it is time to manage the actual pods. Let’s take a look at what we can do.

Looks like a short post to me since there’s onl Get, List and Update

Pod_Get

Just like with site’s the get can be used in conjunction with a podid that might be gotten from somewhere else

This selects the first podid listed when pulling all the pods from all sites and gets the information about that pod. We’ll see the same information when doing a list but just with all pod’s listed.

Pod_List

Those endpoints are the connection servers in the pod. Let’s take a short detour and get the listing for one of those (the podendpoint service only has list and get so you will not see them separately anyway).

It might look lazy to use the select -first one and yes it is a bit but doing  a foreach to explain things also doesn’t really work in my opinion.

Pod_Update

Standard by now, first we need to connect to the podservice.

Under $podhelper we can already see how to set things.

Let’s update  the easy things.

As a result we have updated the name and description of the pod. The other thing we can do is assign the pod to another site. Thankfully I already have two of those created.

Both the lists aren’t required but I added them to show that the pods are spread over both pods now.

 

Creating and managing Sites in a Horizon View Cloud Pod Environment using PowerCLI & Api’s

Intro

Like I said in my previous post about Pod Federations this is a separate post that will show how to handle Sites within a Pod Federation. There are only a couple of API calls that do not include assigning a pod to a site. This is done trough the podservice which I will post about in a next blog post.

Let’s take a look at the site service to see what it actually has in api call’s

So we have Site_create, Site_delete, site_get, site_list and site_update. To Make it myself easy I will use the order of List, create, get, update and delete.

Site_list

With site_list a list of all available site’s will be created, currently I have only one so let’s show that one.

Note a lot of information is shown so let’s take a look at the contents of base and pods.

so again not a lot of information since it only contains a name, description and the pod id’s of the member pods.

Site_Create

Since we already saw in the methods under the siteservice that the create needs a bit more information then just a name let’s take a look again at what is required.

An object is needed of the type vmware.hv.sitebase, we will need to take a look in the API explorer to see what this object should contain. Under Site_create we can click on sitebase.

The sitebase object has 2 properties of which only DisplayName is required. I have tried various ways to keep the description empty but haven’t succeeded so far and with it it the create also doesn’t work so how optional is it?

Let’s create the sitebase object

The $sitebase is not required but shows what the object contains. Now we have enough to create the new site.

Site_Get

In the overview we have seen that a site_get needs a bit more information.

We already know how to get this site id by using site_list, normally you would only use the site_get with an id received from another service like the pod service. For the example I will use the demo site I create in the site_create part of this post.

First I will need to get the siteid

And now we need to apply that to the site_get

Site_Update

As said before for an update method it is better to use the helper service for that service.

now what method’s do we see?

To see the difference between the sitebasehelper and siteinfohelper I will create both objects.

Now let’s compare them.

 

This is again one of those wtf moments, they both do exactly the same! I will use the sitebasehelper for now will update both the Displayname and description. For this I will need to use the getbasehelper 1 step deeper

and apply the update, since neither will generate a response I won’t put any screenshots in.

Now let’s see the result for a site_get for this site now

Site_Delete

We can take a look at it but to delete a site we only need the siteid so let’s remove that site we gave an update.

again no visual feedback but if we do a sitelist there’s only one left.

 

 

Initiating and managing the Podfederation in a Horizon Cloud Pod Architecture using PowerCLI & API’s

One of the new cmdlets for the vmware.hv.helper that I am currently working on is initiating the Cloud Pod Architecture (CPA) and more actions related to this. This blog post will show the basics about initiating, and joining a CPA using the API’s. Doing things with site’s within the CPA will be covered in a later blogpost.

If we look at the services available in the Horizon API’s you’ll see that podfederation is one of them, let’s take a look at that and what method’s are available.

So we can Eject, Get, Initialize, Join, Unintialize, Unjoin and update a podfederation. If we look at the brackets behind the methods than (un)initializing and unjoin don’t need any extra info so let’s get ahead and initialize the podfederation. To show you there’s nothing there yet I made a screenshot of the admin interface.

Initialize the podfederation

Now to initiate the podfederation

And if you are quick enough in switching to the admin interface will also still show it initializing

Get information about the federation

With podfederation_get() we can grab the configuration information.

Not a lot of information but there isn’t a lot more anyway in the podfederation itself.

Join a federation

I have another pod that I want to join to this federation since we’ve already seen that this needs some more input let’s check what it exactly needs.

So we need a remotepod address, presumable one of the connection servers in that pod will be enough, a username where domain\username will do just like in the admin console and a password of the type vmware.hv.securestring. The last one was new for me but thankfully it was described in one of the examples in the api explorer (https://code.vmware.com/apis/75/view and click on Data Object Types).

With this it’s easy to add the local pod to the podfederation

And again if you are fast enough this is also visible in the admin console

And now a get will also show that it has been enabled

Unjoining a Podfederation

If you are braking down a pod because of whatever reason the best way to do this is to unjoin the pod from the federation. As we saw before there’s no extra information need so you just need to connect to a connection server in that pod and do an unjoin.

this is really fast so over several tries I did not succeed in making a screenshot of the admin console.

Ejecting a pod

This is the only podfederation function not available through the admin console as far as I could see. Ejecting a pod for is for me a last option if a datacenter burned down, everything is gone and you want to get rid of the pod. I did it in my lab against an alive pod and had to uninitialize the (now unlinked) podfederation from that pod to be able to rejoin it to the correct pod. This method also requires some input so let’s see what that is.

So we need the podid of the pod to eject, this information can be get trough the pod service

I want to eject the pod from pod2cbr1

and with $pod I can check if I have the correct one

So let’s serve the eviction notice to the pod.

No feedback, nothing but if we check the pod list it’s gone.

I will show how to remove the remnants in the uninitialize chapter.

Updating a Pod Federation

This one sounds bigger then it is since there’s only one thing that we can update in a federation. To do this it is better to use the helper service then to use the podfederation_update method since that can get complicated very fast sometimes. To use the helper service we will need to create some variables first

and when we do a get method on it

After some trial and error I know we need to getdatahelper method to continue

This only show the updates that are currently in the queue to be applied with a get method it’s possible to see what can be set.

What we need to look for is a set so the only options here are setdisplayname that needs a string value and setupdates that needs a load of information and that probably might also be a way to do it but I will use the setdisplayname.

This will give no feedback and nothing will be changed yet, what needs to be done is to apply this update in the helper service to the service.

and if you now do a get on the podfederation it will show the changed name.

Uninitializing a Podfederation

To show the pod uninitialization step I will use the pod that I have ejected from the podfederation pod2cbr1. It is clear that it is a bit wonky if we look at the pod list from that connection server.

So it knows about the pod federation but doesn’t see itself in it anymore.

This is again a fast one so I couldn’t get it visible in the admin console but when checking the data from a get it shows it has been disabled.

Looking from the other pod it still shows the Podfederation as enabled.

No github scripts this time since I will be adding this functionality into the vmware.hv.helper module.

Remove desktop assignment

In this blogpost I showed how to assign a certain vdi machine to a user. This has been made easier since in the vmware.hv.helper module. These days we can use this:

This week I got the logical question from Brandon Smith in the comments about removing the assignment. First I need to re-assign the desktop (I have been building a VMUG presentation about PowerCLI & Horizon view so things got messed up)

the result:

Now what need to be done is setting the base.user to $null. I am going to do this by connecting to the machine service and utilize the machinehelper to update the userdata.

We now need to connect to the machinehelper by doing a read on the machineservice. $machineservice.read will give us the info we need to be able to do this.

From this it becomes clear that we will need the services service and the machineid we want to edit. First let’s put the machine id into a variable.

Now I will read the properties of this machineid and put this into the $machineinfohelper variable

Since I know the user property is under the base we will need to get the base first and then set the user. This done by doing getbasehelper() on the machineinfohelper and then do .setuser(user) on that but let’s see what’s under the getbasehelper first.

A lot of information but as said the one we need is setuser. To assign a desktop we will need to set this to a userorgroupid value (and that is what the vmware.hv.helper cmdlet does). To clear it we will need to set it to $null.

At this point no changes have been made yet! We will need to apply this update first.

And if we look at the horizon console the entitlement has been removed.

And the complete script:

As usual the script can be found on github.

{UPDATE]The Dutch VMUG UserCon was awesome!

Update: the video for my session can be found HERE (it is in Dutch!) or watched down below 🙂

 

I know I used it before but Ooooooooooooh what a rush where the last two days. It’s the day after when I am writing this but like I said on twitter I am still bouncing around after the Dutch VMUG UserCon. My own session went great and hopefully I can put the link to the video here very soon. What I can do is give you the slide deck I used and you can find it at the bottom of this post. The only thing I changed was that I added a translated disclaimer since I had way too much fun creating that one. I did create it as a parody of all the VMworld and .Next disclaimers and because I was in the last timeslot of the day and wanted to give the audience something to laugh when they walked in so they would stay awake after a tiring day at our mini VMworld.

disclaimer

Day one

Now about those two days, at the first day I was one of the three judges in our first (hopefully annual) Hackathon. We had three teams, two from partners: ITQ & PQR and one, that got dragged in by their TAM, from the customer side. PQR build a combination of VRA, home automation and Lego Mindstorms. They gave a voice command to a homey and the Lego Mindstorms robot showed the status of the new vm request while Ikea Tradiri would also show the status with a colored bulb. When a vm was rejected the robot dumped the bit the used as example in the bin aka off the table :D. Team ITQ wanted to deploy a PKS cluster by using giving an Amazon Alexa a voice command that in its turn used a slack channel as messagebus to get things deployed. Not everything worked perfect but in the end they were able to ask Alexa for a Minecraft server and a Steve webinterface would give visual feedback if it worked. As an easteregg Steve could also do a barrel roll. Team Dictu had a different use case, they are busy implementing VRA and their TAM though it was a good idea to get some practice at the Hackathon, they did end up deploying things after some waiting on the slow wifi. For me they were the winners for trying to do something almost new to them and getting results as well. That is why I gave them my personal spot price: a six-pack of Achelse Kluis Trappist beer. Overall, we decided not to have a real winner but everyone won as it should be. The theme of Fun, beer and Gyros fitted perfectly and together with everyone who was there I am sure we had an awesome time. Just next year I might think about creating a non-company but vCommunity team of all sorts to do some EUC stuff.

The swag table had some interesting stickers at the hackathon

During the day at day one there were also several VMware R&D sessions (NDA I believe) and a VCDX workshop. The R&D Sessions could use some more audience but the vcdx workshop was very full I believe. For me day one was already a huge success while day two was still coming.

Day two

Day two or more actually UserCon day. I must admit that I didn’t attend a lot of sessions but the keynote by no-one else then Pat Gelsinger was great. He did not get a lot of questions during the q&a part but it filled the time and he had to use the I cannot confirm nor deny that several times. Some of the sessions I attended where a NSX jumpstart for VDI by Pascal van de Bor, a security awareness session (marketed as security for EUC admins & Consultants) and some others. At 15.50h it was time for my own session that was in fact an introduction to PowerCLI for Horizon View. I managed to spoon-feed the audience 61 slides in a bit over 20 minutes without them even noticing it. I had 16 slides with static content including vmug, title, introduction and overview pages. The rest where simply moving gifs (without loops!) that I recorded using ScreentoGif. This allowed me to demo the functions and usage of the results without boring the audience with 10 minutes of typing errors by doing it manually. And yes, some things I had to do over 10 times to get it rights so it was worth doing it that way. I managed to get myself in a relaxed state by listening to some music (Katie Melua this time) while going through the deck one last time. During the presentation I hardly felt the nerves and even managed to keep my cool and not speak too fast like I did before. While last year with Hans it was a success this time was even better and I was really satisfied with the results.

You can find the final presentation here: https://wkursten.stackstorage.com/s/Y8hkexL8qDTXVVb it was built in Dutch but most are demos anyway.

I think these are my lucky socks now since I wore them during day 2.

 

Multi vlan Network for Horizon View using PowerCLI & API’s

One of the things I wanted to do for a while is to write an API version on how to use multiple dvSwitch portgroups with Horizon View linked clones. With instant clones there’s a gui way to select multiple portgroups but for instant clones the only was to do this was to use the View PowerCLI. This gets installed with the connection server and can only be used from there. What you do is create a file, edit and apply it. Johan has described this process very well on his blog. I decided there had to be a way to do this as well with ‘regular’ PowerCLI & the api’s.

The api explorer shows a property named networklabel for both desktop pools and rds farms. This entry showed me what data I needed to configure. I spent most of my time in gathering all the data for this. As you can see in the script I had to dig rather deep to get all information like hostorclusterid and snapshotid. This information then needs to be put into an object called nics.

The script I made is a working prove of concept and doesn’t contain logic about what portgroups to apply. It just grabs all portgroups that comply with a simple filter. It then grabs the id’s for those and configures them to use for the pool. The script grabs information using the snapshotid but in my testing it’s 100% safe to change snapshots or golden images after that, is just uses that information to know where to configure things.

Something I found during testing is that the maximum amount of labels is respected and spread over all port groups as long as there are labels available. If the system runs out of labels it will continue using only the last configured label! I have tested this on View 6.2 and 7.3.2 with vSphere 6.5 on both methods of configuring the portgroups.

This is the script, it asks for some required information at first. This way you don’t have to put a password in plain text in the script. You can see I have the maxlabeltype and enabled properties pre-configured as LIMITED and $true. If the maxlabeltype is UNLIMITED the composer would stop using any other labels configured after that one and if enabled would be $false that label wouldn’t be used at all..

I used a lot of variables and arrays with the names as they are pulled from the data, that explains their long names. Afterwards it doesn’t give any feedback. For this I created a separate script so you can separately check what is configured before or after you change the configuration:

And the result:

In the end the script looks and is way more complex than the ‘old’ way to assign multiple vlans. On the other hand it is way more flexible to use in any scripting you are already using for the automation of your Horizon environment.

As always both scripts can be found on Github here and here.

Finding Horizon View local entitlements using PowerCLI

Intro

In a previous post i mentioned that finding the entitlements for a user from the Horizon side of things can be a bit of a hassle. If only active directory groups are used its dead easy: just use the Active directory commands for those groups. If the groups are used for multiple pools and if you have assigned desktops things get a bit more complicated. For now I will only concentrate on the local pod without global entitlements.

getting that info

To get started the vmware.hv.helper module has the get-hventitlement command. As almost always a very useful one but it has some flaws. First it requires full domainname\username or username@fulldomainname.

For example

or

Both work but

gives this message: Get-HVEntitlement: No entitlements found with given search parameters.

At least

If you add the -type group to this command you get all group entitlements

gives an error message that the -user argument does not match the “^.+?[@\\].+?$” pattern. With this last one you at least get an error so you know where to look but not displaying any entitlements is an issue for me.

So, back to the results of these commands, I have assigned the user user1 the following rights

  • Pool04 directly and by using a group
  • directly on a single desktop in pool04.
  • Pool01 only by group.
  • Paint rds app by group
  • Calculator rds app direct
  • Wordpad rds app by both group & directly

When using the get-hventitlement without anything else it doesn’t seem to show a lot of usable things

If you put this between brackets followed by a period and one of the properties a bit more info is shown.

Some information about the user, not very usable the session data property gives some information about current sessions (none at the moment)

With the localdata property it looks like we hit the motherload jackpot thingy

Very good, a lot of id’s so what can we do with those? For now I will put this into $entitledids.

I read something about get-hvinternalname when checking out the module, sounds usable.

Ah, so this needs an entityid as input, a machine is an entity so let’s try it. This might need a foreach though because the output gave machines and not machine.

Damn, that’s not usable, let’s double-check with the other id types

The ones we need are readable, couple of them not but I don’t those will be missed.

The missing machine name is actually easy to solve by doing an api call:

Conclusion

Because this is rather easy to use and since I didn’t have a direct use case for that I decided not to create a complete script. With get-hventitlement, get-hvinternalname and maybe an api call here or there it’s very easy to pull the information about which account or groups have what rights. To see if a user belongs to a group can easily be done with any of the multitude of scripts for that here’s a good example of those.

 

Removing faulty Horizon desktops using PowerCLI

So last week there where a couple of posts on vmtn about people wanting to automatically removing or refreshing faulty Horizon desktops. With faulty I mean desktops in Agent Unreachable or in error state or whatever status are available. Since this was something i had been investigating anyway I decided to make a script for it that had separate menu’s for the status the desktop needs to be and to pick the desktop to be deleted. The latter part can be rebuild to do all those desktops at once  in case something breaks pretty badly during a recompose of the pool.

The largest part of the script is for creating the menu’s. Since the amount of returned desktops is variable and names differ it’s not possible to use a static menu. Instead I have used a menu structure created by Roman Gelman and that can be found inside this script on github. The part that gets things done i have listed below. The $spec array doesn’t need to be created but it is required in the API call to remove the desktop, Powershell assumes everything true by default when it’s empty but it just has to be called otherwise you will get a big fat red error. To remove multiple desktops at once machine_deletemachines needs to be used with an array filled with desktop id’s and $spec.

As always the complete script can be found at Github where it will also be updated. This is how it looks in the end:

Update

After the comment below I decided to create the script to delete all desktops in a certain state. It’s a variation of the script above, just a bit shorter. Again it can be found on Github. Please be aware that due to a limitation in get-hvmachine both these scripts will only handle 1000 desktops at a time. It is safe to just repeat the script to do the rest.

https://github.com/Magneet/Various_Scripts/blob/master/remove_faulty_VDI_desktop.ps1

https://github.com/Magneet/Various_Scripts/blob/master/remove_multiple_faulty_VDI_desktops.ps1

Horizon view vCheck : Pool Overview plugin

So one of the things still missing in the Horizon View vCheck was a plugin that simply gives an overview of all Pools with their status. In short what I am talking about is a translation from this view:

Although this sounds easy there where a lot of challenges for this one. First of all there are three separate pool types: Automated,Manual and RDS and all of them have subtypes like VIEW_COMPOSER,VIRTUAL_CENTER,FULL_CLONES,INSTANT_CLONE_ENGINE,UNMANAGED or RDS and not all of these subtypes are available for all pool types. This gives a lot of options that need to be separated for the pool types. Also the VIRTUAL_CENTER subtype is used for both manually added desktops that reside on a vSphere environment and for an automatic pool creating full clones. The FULL_CLONES subtype I haven’t been able to create in my lab yet.

Further outputs like true, false or any of the subtypes above weren’t clear enough for me to use as output. For this I learned a new trick in my book called switch.

Some documentation for the switch command can be found here but what it in short does is match the variable u use as input and sets or gives some output based on that. Also it can do a comparison as in above example so I was able to distinguish between Full Clones and Manually Added vCenter Managed Desktops. One thing to be aware of is that it will go trough the complete list. At first I had the two lines with the comparison in it at the top but that got overwritten since below it VIRTUAL_CENTER was recognized and the $sourceoutput would be based on that.

The Automated and Manual pools use a very similar set of code, the biggest difference is that one gets the data from the AutomatedDesktopData propertywhile the other gets it from the manualdesktopdata property.

The RDS block gives a totally different view though. The information had to be pulled from the farms that are the backend for the desktops.

And when done I ended up with the following script. As usual it might get some improvements or I need to squash some bug so better check the latest version on Github.

And a screenshot of the result:

Pulling Horizon View events with PowerCLI in a nicer way

So in one of the first posts I did about Powercli 6.5 was a bit about pulling events from the Horizon View events database. This off course was not the crude way to pull the events. Another was was using the event dumper fling but that one is a bit to slow for my taste. I decided to create a script that leveraged and had the speed of PowerCli but actually gave some options and that dumped the results in a .csv file since reading them from the command line will be too much at some point.

What I ended up with is a 71 line script where 18 are informational, 8 blanks, 1 that pulls the information and the rest for the menus and error handling. I decided on not giving any real errors when things don’t work but just red text. A lot is done with try <> catch and a if elseif else.

There is no need to load any modules but the scripts assumes at least PowerCLI 6.5 with the vmware.hv.helper module loaded.

I have created a github repo for these smaller scripts over here.

And how does this look in the end??