Here is a network diagram of the connection:

The Cisco router is the only device we fully control in this configuration. The ADSL modem is configured and managed by our ISP. The Cisco router is a 2651XM running IOS 12.2 – not the latest and greates router, but it works – most of the time.

Today someone was clogging the network…

Troubleshooting

I tend to use Googles’ DNS server, 8.8.8.8, to verify my Internet connection. I have yet to experience that this server is down. Running a ping to this server indicated a very high latency.

jahansen@H720:~$ ping 8.8.8.8
PING 8.8.8.8 (8.8.8.8) 56(84) bytes of data.
64 bytes from 8.8.8.8: icmp_req=2 ttl=47 time=820 ms
64 bytes from 8.8.8.8: icmp_req=3 ttl=47 time=974 ms
64 bytes from 8.8.8.8: icmp_req=4 ttl=47 time=1100 ms
64 bytes from 8.8.8.8: icmp_req=5 ttl=47 time=741 ms

The latency is normally somewhere between 20-50ms. Which device is adding all this latency? A traceroute (tracepath/tracert) to the same server revealed the where:

jahansen@H720:~$ tracepath -n 8.8.8.8
 1: 192.168.0.177     0.122ms   pmtu 1500
 1: 192.168.0.1       2.300ms 
 1: 192.168.0.1       2.363ms 
 2: 2x.1x.1x.x3       5.272ms 
 3: 2x.1x.1x.x3       3.521ms   pmtu 1492
 3: x0.x2.x3.1        854.072ms 
 4: x4.x8.x4.61       940.408ms   asymm 6 
 5: x4.x8.x3.18       942.447ms

When looking at the latency of the traceroute, no significant (on this network) latency is added from my client to the Cisco router. The same goes for the Link network between our router and the ISP ADSL modem, but look at the next line. My data is having a really hard time getting from the modem to the next router adding >850ms latency. No wonder the Internet connection appears to have an error.

The added latency can be a result of multiple issues, but in this case; Some technical issue in the network outside this building or someone inside is clogging the line making the queue to build up.

I did a quick Google search to find the commands and it appears to be really simple to “monitor” clients on the inside. It is done by enabling flow monitoring on the inside interface of the NAT. In this case, the FastEthernet 0/1 is the NAT inside.

router#conf t
router(config)# interface FastEthernet 0/1   #Inside interface of the NAT
router(config-if)# ip route-cache flow 

Once ip route-cache flow is enabled, give it 10 sec or something to get the counters running. Now, to list all clients and the number of packets they have transmitted/received

router(config-if)# do show ip cache flow 

The size of the list depends on who many connected clients. Pay attention to the last column; Pkts

SrcIf  SrcIPaddress   DstIf   DstIPaddress  Pr  SrcP  DstP  Pkts
Fa0/1  192.168.0.217  Fa0/0   1x.16.1x8.1   11  FCD8  00A1  3 
Fa0/1  192.168.0.177  Fa0/0   4x.x9.x2.179  11  C542  1A26  197 
Fa0/1  192.168.0.226  Fa0/0   1x3.1x4.x32.7 06  E5D8  01BB  1 
Fa0/1  192.168.0.177  Fa0/0   1x.x8.11x.4   06  DF52  0050  15K

In this case, the number of packets the devices with the 177 IP address is increasing significantly faster than other devices – and 177 happens to be my IP address… Wooops.

The case with the 177 address being the sinner is just an example. In the real case someone was uploading a large file to Dropbox. The ADSL upload speed is somewhat limited and the result being very poor performance both ways.

Have a look at the Wiki pages for the short version.

For this post I preparing a Windows XP VM with 20GB c: drive and 10GB e: drive and I’m going to use the e: drive in this example. Before adding any data to the VM is using <4GB disk even though it has 30GB provisioned. (The additional GB is the VM swap file).

Lets start adding data.

I have added two iso files to the e: drive. As we can see the used space has increased accordingly.

Lets delete one of the ISO files and see what happens.

Yeah, the e: drive got some more breathing space, but the used storage is still the exact same as the one about. Why is that? Well, the same applies here as with the Solaris JumpStart server; When a file is deleted from the file system, the blocks on the disk still contains data, but the pointer is removed. This is why it is possible to recover deleted files from a hard drive. To “free” the wasted blocks, we need to zeroize the blocks.

Many suggest using SDelete from Sysinternals (Microsoft) to do this. It is a command line tool that works perfectly, if you’re zeroizing the c: drive. In my case I’m trying to zeroize my e: drive and there is a bug in SDelete 1.6 that ignores the driver letter parameter. On one of my templates I did this using SDelete and it just works. I observed that a file was created directly on c:\ that got very large during run and got deleted at the end, pretty much the same as doing a dd from /dev/zero.

Due to the bug in SDelete, I decided to try dd for Windows from chrysocome.net instead. It works just like dd on unix using virtual devices for /dev/zero since that device does not exist on Windows.

I decided to go with version 0.5. Downloaded, extracted to c:\ and executed the dd command to fill my e: drive with zeroes.

The e: drive contains a file full of zeroes and the operating system reports that the drive is full. Zeroes in a file is valid data as well, so to free the wasted space from the OS, delete the file.

c:\dd-05\>del e:\dd.file

The OS reports that there is available disk space on e: now, but on the datastore the VM is  still using the blocks. Use vmkfstools free the zeroized blocks on the datastore. Before executing vmkfstools the VM have to be shut down. Power off the the VM and open a shell on the ESXi host. I’m using ssh. I added an additional disk to show this procedure and I need to identify this disk.

The .vmdk file is the configuration file for the virtual disk and the -flat.vmdk is the actual file. Execute the vmkfstools command with the -K option and provide the full path to the .vmdk file as parameter.

~# vmkfstools -K /vmfs/volumes/Local_SSD_Dell/xp-x86-relcaim/xp-x86-reclaim_1.vmdk

And voila, have a look at that! I’m down to approx 6GB again which is the c: drive and one .iso file on the e: drive.

This is how the wasted thin provisioned datastore usage can be reclaimed on a Windows based virtual machine. I have done this on both Windows XP and Windows 7 with the same result.

Good luck reclaiming disk space.

Create a new virtual machine with a single hard disk of 20GB. Independent of which disk type chosen, the OS will see a 20GB disk. A thick provisioned disk will use 20GB on the storage device, while a thin provisioned will increase over time. After installing the operating system and a few apps, the VM is using about 5GB on the datastore (Thin-25). Then copy a 5GB ISO file to the VM and it bumps up to 10GB datastore usage (Thin-50) and so on. This is where VMware distinguish between provisioned and used disk space when talking about resources.

I’m using thin provisioning all the time and I haven’t had any issues with it, but keep an eye on the free space on the datastore – you really don’t want to see 0kB there…

I said that thin provisioning is a feature where the disk file increase over time. When a file is deleted from the virtual machine OS the free space is available for the OS, but disk file on the datastore is not reduced and will keep using those blocks.

How can wasted and unused thin provisioned capacity be reclaimed? 

For many of you this is old news and if you google it you’ll find this;

Do a storage vMotion of the virtual machine to another datastore to reclaim unused blocks.

This is the method I’ve been using the last few years as well and it works with vSphere 4.x, but it will not reclaim blocks for you on vSphere 5. The reason is that VMware introduced a new method when moving data. When doing a storage vMotion to a datastore with different block size the legacy datamover, fsdm, is used and blocks are reclaimed, but when moving to a datastore with the same block size the datamover introduced in vSphere 5, fs3dm, is used. Since VMFS5 is using only one block size, you might not have any datastores with different block size. As a result, the “work-around” to reclaim blocks using storage vMotion doesn’t work any more.

Thanks to Duncan Epping at yellow-bricks.com for enlightening me on the block size impact and helping me understand why the storage vMotion work-around doesn’t work anymore.

Solaris JumpStart Case

Some say that reclaming disk space is a waste of time, but I had a case a couple of days back where I really wanted to. I have a Solaris JumpStart server VM that is running Solaris 10 on x86 and contains a flar image for a Sun SPARC servers. The JumpStart server is used to deploy Solaris operating system on physical hardware much like Windows Deployment Services is for Windows operating systems. This JumpStart server is exported to a Virtual Appliance and sent to multiple locations where the flar image is to be deployed. The VM has a single virtual disk of 100GB.

The Solaris operating system is using about 5GB (yes, a full install) and the flar image is about 7GB, plus some other stuff so when running df -h on the machine, it is using about 15GB. The virtual machine disk is thin provisioned and once exported I expected the size of the virtual appliance to be something similar. They are usually smaller due to compression, but the result was 24GB and I thought; “ahh, of course. At one point I did have 3 different flar images on the JumpStart server. I deleted two of them, so if I do a storage vMotion before exporting I’m all good”. Guess what, the appliance did not get any smaller.

In order to reclaim wasted disk space, the blocks available to the OS must be zeroized followed by some vmkfstools magic.

Unix/Linux

The jumpstart server virtual appliance is an ova file of about 24GB which contains the Solaris OS and a single flar image. When the appliance is deployed it is actually using approx 35GB.

Looking at the disk usage, the OS isn’t even aware that there are 24GB of used blocks, but the virtual machine disk file says otherwise. From the OS perspective, the partition where the flar image is stored, /export, is only using about 11GB. 

I know this partition have been using much more at one point, so I’m gonna fill this partition with zeroes using dd. Since this VM has a 100GB disk, filling it with zeroes will take some time and be aware; the VM will require <100GB of available storage on the datastore.

root@jumpstart:/$dd if=/dev/zero of=/export/dd.file
write: No space left on device
166428705+0 records in
166428705+0 records out
root@jumpstart:/$

The VM is now using almost all of the provisioned disk space, both from the vSphere and the OS perspective.

The /export partition contains a file full of zeroes and the operating system reports that the drive is full. A file containing only zeroes is a valid file as any other, so to free the wasted space from the OS, delete the file.

root@jumpstart:/$rm /export/dd.file
root@jumpstart:/$df -h
Filesystem        size used avail capacity Mounted on
.
/dev/dsk/c1t0d0s7 90G  11G  78G   13%     /export
root@jumpstart:/$

The OS reports that there is available disk space on /export, but the VM is still using the blocks on the datastore. Free the zeroized blocks on the datastore using vmkfstools available on the ESXi host. Before executing vmkfstools the VM have to be shut down. Power off the the VM and open a shell on the ESXi host. I’m using ssh. Execute the vmkfstools command with the -K option and provide the full path to the .vmdk file as parameter. (Sorry about the space in the path. My bad)

vmkfstools -K /vmfs/volumes/Local_SSD_Dell/Jumpstart\ Server/Jumpstart\ Server.vmdk

And voila, have a look at that! The VM is using approx 24GB on the datastore.

Exporting this “reclaimed” virtual machine to an .ova virtual appliance reduced the footprint of the appliance from approx 24GB to about 15GB! That’s more like it. In addition to reduction of the virtual appliance, the storage requirement when deploying has been reduced from 36GB to 25GB.

This is how the wasted thin provisioned datastore usage can be reclaimed on a Unix/Linux virtual machine and in this case it reduced the download size of the JumpStart server virtual appliance with almost 40%.

Maybe I’ll do one for Windows as well…

I tried to find a guide using google to do this, but without any luck. Lets have a look at how I changed the vCenter server managing the desktop pools.

Disclaimer: The method described here are not confirmed or documented as a supported method by VMware and I take no responsibility for your actions.

VMware View – Change vCenter server on Pools

When deploying a virtual desktop pool, one of the initial settings or choices we have is which vCenter server we want to use. VMware View Manager can utilize multiple vCenter servers, but a desktop pool can only be pointing to a single vCenter server. The VMware View infrastructure running in this environment is VMware View 5.0 Permier, but non of the pools are linked clones. This is also one of the reasons we can migrate to the vCSA and eliminating the Windows server license.

Just a note; Previously the View Composer service had to be installed on the vCenter server – this does not apply any more. If you are using linked clone pools, you can still use the vCSA for the vCenter server service, but VMware View composer still have to be installed on a Windows server. I’m guessing that we’ll see the View composer service as a add-on to vCSA a few versions up the road.

In this environment, re-creating all virtual desktops weren’t really an option. Users were still logged in at their desktops, some of them with active/connected session and some with disconnected sessions. Reboot is also something the users dislike – bloody developers! Those demanding buggers…

Ok, so looking at the View Manager page, all pools seems to be ok even though the vCenter server defined on each pool was wrong. Clicking through the pools, no errors were listed at the pool or desktop lists. If I had tried to provision a new desktop by increasning one of the existing pools, it would fail because the defined vCenter server is not connected to any ESXi hosts, templates or even virtual machines.

We do want to add the vCSA to the View Manager to generate an ID for the vCSA in the View LDAP directory or “database”. This is done at the View configuration -> Servers -> vCenter server tab. Add the vCSA as any other vCenter server.

Luckily, the pool settings are defined in the View LDAP and these settings can be exported. The View LDAP is running on the connection server and can be exported by executing a vdmexport command on the server. Open a command promt or the VMware PowerCLI Powershell promt and execute the following;

PS C:\Pro..\VMw..\..View\Server\bin> vdmexport -f c:\ldap-export.ldif

Note: On View 5.0 the export will not be sealed and fully readable. On View 5.1, the vmdexport default to a sealed, non-readable file. Execute the vdmexport command with the -v option to get a readable file.

This will create a ldif file on c: named ldap-export.ldif which contains all configuration data for the respective View connection server. This is the file I edited to modify the pointer to vCenter from the old vc01 to the new vcsa01.

I first had to identify the ID of the legacy vCenter server and the vCSA by opening the ldif file in notepad++ and searching for the hostname. The IDs are fictious to make it easier to show the difference and modifications to be made.

The legacy vCenter server:

dn: CN=abcd1234-49cb-4fce-ae39-5dbcba9120d1,OU=VirtualCenter,OU=Properties,DC=vdi,DC=vmware,DC=int
changetype: add
objectClass: pae-PropertyObject
objectClass: pae-VirtualCenter
cn: abcd1234-49cb-4fce-ae39-5dbcba9120d1
pae-VCCreateRampFactor: 20
pae-VCUserName: root
pae-SVIUserPassword: {SSO-AES:1}ir+/2x2x2x2x2x2x2xx2x2x2x
pae-VCURL: https://vc01.v12n.local:443/sdk

This is the vCSA:

dn: CN=efgh5678-49cb-4fce-ae39-5dbcba9120d1,OU=VirtualCenter,OU=Properties,DC=vdi,DC=vmware,DC=int
changetype: add
objectClass: pae-PropertyObject
objectClass: pae-VirtualCenter
cn: efgh5678-49cb-4fce-ae39-5dbcba9120d1
pae-VCCreateRampFactor: 20
pae-VCUserName: root
pae-SVIUserPassword: {SSO-AES:1}ir+/2x2x2x2x2x2x2xx2x2x2x
pae-VCURL: https://vcsa01.v12n.local:443/sdk

The dn value is the ID of the vCenter server, in this case being the vc01.v12n.local and vcsa01.v12n.local.

Now, I had to find the pool definition and replace the defined vCenter server ID with the new one. By searching for the ID of the legacy vCenter server, I found a section looking something like this;

dn: CN=win7-auto-full,OU=Server Groups,DC=vdi,DC=vmware,DC=int
changetype: modify
add: pae-VCDN
pae-VCDN: CN=abcd1234-49cb-4fce-ae39-5dbcba9120d1,OU=VirtualCenter,OU=Properties,DC=vdi,DC=vmware,DC=int
-

The pae-VCDN is the value I modified. I replaced this VCDN ID with the on for the vCSA listed above:

dn: CN=win7-auto-full,OU=Server Groups,DC=vdi,DC=vmware,DC=int
changetype: modify
add: pae-VCDN
pae-VCDN: CN=efgh5678-49cb-4fce-ae39-5dbcba9120d1,OU=VirtualCenter,OU=Properties,DC=vdi,DC=vmware,DC=int
-

I repeated this operation for all the desktop pools administrated by the connection server. Do not execute a replace all in this document! Doing so will break the ldif file and you do not want to import a broken ldif file, trust me. I tried and ended up re-installing the View Connection server and import the initial exported ldap-export.ldif file. This was my first complete restore of a View connection server and I’m quite impressed how easy it was.

Once all pae-VCDN for all pools have been changed, save the exported file as a new file in case something goes wrong and we need to revert our changes. I saved it as ldap-export-modified.ldif.

To load the changes on the View connection server, open a command prompt or execute the Power CLI shell again and execute the following:

PS C:\Pro..\VMw..\..View\Server\bin> vdmimport -f c:\ldap-export.ldif

The vdmimport command will only inject modified settings to the LDAP directory and leave the others as they are. In this case, the only modification made is the pae-VCDN on the pools. Looking at the pool list in View Manager now, all pools are now pointing to the vCSA.

The View connection server service did not have to be restarted for the pool list to be updated – a simple refresh did the trick. I’m not sure if I restarted the View connection server service before increasing one of the pools. At first it failed due to missing customization specification defined in the pool settings, but a simple export – import of the specifications did the trick.

Right click the specification to export and use the import button in the menu bar to import on another server.

The infrastructure has been upgraded from ESXi 5.0 and vCenter 5.0 to ESXi 5.1 and vCSA  5.1.0. The VMware View desktop pools have successfully been migrated from the legacy vCenter server to the brand new vCenter Server Appliance. Remember the demanding developers? Guess what; non of the virtual desktops had to be rebooted or re-provisioned during this upgrade. I do have to reboot when upgrading VMware tools though, but that’s another story…

This is a fairly easy transition in a virtual environment with servers only, but adding virtual desktops and VMware View to the same infrastructure complicates things. I tried googling and reading VMware documentation, but I could not find any supported method to migrate virtual desktops pools from one vCenter server to another, which in this case is the new vCenter Server Appliance.

Disclaimer: I must emphazise that the method described here are not confirmed or documented as a supported method by VMware and I take no responsibility for your actions.

I’ll take you through my upgrade from start to end starting with the vCenter server “upgrade”

The enviroment we’re talking about here is rather small. It consists of a single vCenter server, three ESXi hosts, one VMware View Connection server and a combination of automated and manual desktop pools. Non of the pools are linked clones pools provisioned using VMware View Composer, only full cloned virtual desktops desktops, both manual and automated.

Deploy vCenter server appliance

The first step in the upgrade phase is to deploy the vCenter server appliance. Deploy as any other virtual appliance and define static IP adressing during deployment. Once deployed, power on and log in to the appliance management page at https://vcsa01:5480. You’ll meet a configuration wizard page. You want to cancel this one to configure hostname and regenerate certificates before running the wizard.

In this case I chose to use the embedded database, which is based on Postgres in the latest vCSA release. VMware states that the supported number of hosts are 5 and only 50 virtual machines. Not sure why they’ve set the numbers this low though. Since the only external database supported in this release is Oracle and the current vCenter server is using MSSQL, I had to re-create the whole folder and permission structure before adding the ESXi hosts.

Adding ESXi hosts

The vCSA is configured with datacenter, cluster, folders and permissions. I consider vCenter to be a non-critical service and since disconnecting an ESXi host from vCenter doesn’t affect the running virtual machines, I don’t see any issues doing it. (I did this after the developers went home for the day). Before disconnecting, I disabled both VMware HA and DRS. When disconnecting ESXi hosts from vCenter, the list of virtual machines will not be pretty – they will be gray and with the status of disconnected. Don’t worry, they are still running.

On the vCSA, I added all three hosts to the new cluster using the FQHN. A reminder; the Datacenter name, cluster name and folder structure is identical between the vCenter and the vCSA server. All virtual machines will initially be added to the Discovered Virtual Machines folder so we’re looking at a drag and drop session here, putting all machines in their respective folder.

The Windows based vCenter server can be shut down and the vCSA is now managing all ESXi hosts and virtual machines in my environment.

ESXi upgrade

In this environment, running only 3 ESXi hosts, manual upgrade of the ESXi hosts are very easy. Starting with the first host, migrate all virtual machines using vMotion to any of the other two hosts in the cluster and put the host in maintenance mode. If DRS is enabled, it will vMotion machines for you when activating maintenance mode.

It’s fun doing vMotion on 10GbE!

SSH shell access is enabled on all hosts and the ESXi upgrade zip file uploaded to one of the shared datastores. Log in to the first ESXi host and execute the following command to start the upgrade:

~# esxcli software vib install -d /vmfs/volumes/iSCSI_01/updates/ESXi510-201210001.zip

This will start the upgrade and once completed successfully, reboot the host to load the new VMkernel version. When the host is back online, exit maintenance mode and repeat the task for additional hosts.

~ # esxcli system version get
 Product: VMware ESXi
 Version: 5.1.0
 Build: Releasebuild-838463
 Update: 0

For some reason, in my case, vMotion was disabled on my vMotion VMkernel port. I’m not sure why, but I just enabled vMotion again on the VMkernel port and it was all good.

The virtual infrastucture is now running on vSphere 5.1.0 ESXi hosts and the vCenter server appliance, vCSA, is running version 5.1.0a.

The morning after the VMworld Party, they got to have something exciting to present for people to get out of bed. The headline of keynote was Genius Machines.

Kevin Slavin was talking about how algorithms shape our world. Kind of interesting to see how algorithms and computer systems affect how and where we build infrastructure. You can find the presentation on youtube.

The second speaker was Dr. Dennis Hong displayed some really impressive robots, Darwin and Charli. These are human like robots that walk. What? Is that all? Yes, more or less. They know how to walk. It appears to be really difficult to reproduce the human way of moving around into a machine. Have a look at www.romela.org for more info and videos.

The third speaker was Chris Urmson, the leader of Google Self-Driving car project. This is amazing stuff. Using all kinds of sensors; radar, laser, etc, Google has managed to get this car driving around without anyone operating the wheel. It uses its sensors to create a precise image of the topology around the car and make decisions based on that. Google it and you’ll find some videos on youtube.

Okay, now over to my lack of planning skills.

I’ve attended VMworld since the first VMworld Europe was hosted in Cannes in 2008. Back then I was new to virtualization so I went to a lot of sessions and talked to experts at the expert bar. The next year I spend more time with labs and the solutions exchange, which I think is one of the main reasons for being at VMworld. The breakout sessions will give you many hints and interesting hardware or software that you might want to check out. If the vendor is present at the solutions exchange, and they normally are, you can get first hand information about the product right away and get some contact info for later.

Here is what I do; I spend a lot of time with breakout sessions the first days, mixed with labs to get the latest and greatest. This year I decided to spend day 4 in the solutions exchange. I left the keynote and went over to the solutions exchange. This is why I suck at planning.

I thought; it might be that the doors don’t open until later. I checked the agenda, and guess what; no solutions exchange on thursday! WHAT?! I have a whole list of stuff I wanted to check out in threre. I cannot remember it being closed on thursdag before. So for all guys out there; have a look at the agenda at a glance if you are attending the Europe conference in October.

VMware View and Storage: Thinking outside the box (of disks)

The last few years I’ve been looking for the breakout session that Jeff Whitman and Jim Yanik (both VMware) are presenting. They always manage to give the little extra and it does not matter what other sessions are running in parallel. This year they talked about utilizing flash in a VMware View environment and the brought in Victor Backman from Fusion-IO to assist.

My notes from the session:

• View 5.1 storage features
  • Customize drive letter of disposable disk
  • support for 32 host cluster with nfs
  • View storage accelerator
  • In memory cache of common block reads
  • View Composer Array Integration
  • Utilize storage array in the same way as VAAI, but for composer.
• Test configuration
  • AMD processors, 256GB, FusionIO
  • View Planner
  • Allocate more memory to the host – might fail, not pretty
• Sizing the Storage
  • myvirtualcloud.net – storage planning tool
  • Replica – thing provisioning
  • Memory reservation to eliminate swap file, save space on the storage
• Benefits
  • Simplicity
  • Scaling, add another card or add another host
  • Only relevant for Floating Linked Clones
• Cost
  • 17k USD for the whitebox server.
  • 175 desktops pr setup
  • 26k for a Dell with the same configuration
• Demo; Single box, 500desktops,
  • ioSphere – management of the card
  • Really cool demo – login view, open a document and save

This presentation concluded my VMworld 2012 in San Francisco. I have three things I’d like to point out.

• Make the hands on labs work. That’s one of the main reasons for me being here.
• Figure out how to provide Internet to 40k+ devices
• Leave the Solutions Exchange open on the last day as well (probably costly, but)

It has still been a good show. I’ve learned about new products, upgrades and products that can help to succeed with getting to the 90+% goal that the (almost) CEO of VMware, Pat Gelsinger, stated in the keynote the first day.

Hope to be back in San Francisco next year!

Today I spent my time in labs and and a few sessions.

Build and operate the business desktop (HOL-EUC-02)

This lab presented some more details on how to administrate and configure the virtual desktop infrastructure using group policies. VMware provide ADM files that can be used in group policy objects to configure both the View agent and View client on a desktop. This simplifies management big time. I’m not going to list all here, but a “note to self” list:

• Disable time zone forwarding
• USB authorization – look at linux-usb.org/usb.ids for (exactly) USB ids.
• Clipboard configuration can be set to one way only securing data not being extracted from the virtual desktop.

One thing that goes hand in hand with Steve Herrod and keynote; Demo. Horizon and VMware View, combined with Mirage technology, demonstrated how it is possible to get a user upgraded from Windows XP to Windows 7 and still keeping profile and data more or less live, getting the same data and apps on an iPad and on a Android device. It was a really good display of how powerful Horizon and View is combined. It think we’re going to see a significant increase in the use of these products in the next 6 months.

For me, day 2 was the great lab day.

Build and operate the Mobile Secure Desktop (HOL-EUC-01)

This lab gave a brief introduction to some a few features in VMware View and vShield and how they can be utilized.

• Patch all your desktops and recompose pools to make sure all machines are patched.
• Utilize vShield to prevent machines that are not suppose to communicate to communicate with each other. A very easy; Machine 1 should not be able to communicate with machine 2.
• Add vShield Endpoint for additional security.

Deliver Optimal Performance with VMware vSphere 5.1

This lab was more advanced, and performance tuning and optimization is often more advanced. A whole lot of reading in this lab and here are a few notes from the lab:

• Kernel latency above 2ms indicate storage problem
• Use the paravirtualized drivers for better performance and reduced CPU usage.
• RDMs and VMFS have the same performance so why to RDMs?
• VMware vSphere 5 Memory Management and Monitoring diagram – KB2017642

Java applications are a challenge in a virtualized environment, due to memory management. Java has its own memory management, garbage collector, and if the VM starts ballooning, the Java app might end up running really really slow. During the lab i was using the vSphere web client and while looking at Java memory management and how to solve issues regarding memory usage, this message pops up;

That is just great!

vCOPS manager for Enterprise (HOL-EUC-05)

The vCenter Operations Manager is a monitoring tools that really helps you monitor your virtual environment and points out potential issues for you.

Have you ever seen this woman?

Well, I met her in the lab here. She complained about her newly p2v’ed developer machine and blamed the virtual infrastructure. Using vCOPS manager, the lab guided me through how to find the real issue and how to send the problem back to the developer. Isn’t that cool? A tool that helps you blame the one who complained in the first place?

Horizon Application Manager (HOL-EUC-04)

In the End user computing section, everyone is talking about Horizon and I must admit; I understand why. This is a very exciting product that, as I’ve said earlier, we’re going to see an increase in deployments very soon. We just need to get going with VDI first.

About the lab.. Ehm.. I think I’ve seen this before? It is more or less the same lab i took last year and if I’m not totally wrong I said; This is a good introduction to Horizon. And it is.

PCoIP – Performance deep dive for successfully optimize and customize VDI

I did go to a session regarding PCoIP today. When I scheduled my VMworld, I select Advanced Technical sessions. I have to say, today was the first real Advanced Technical one. The guys on stage really put up a show, a technical one that is. Presenting and demonstrating what the different values of the PCoIP variables actually does and what you can achieve by tuning them. A really good presentation, so for you guys going to VMworld Europe; Look for Breakout Session EUC2620. I hope the guys get the chance to go to Europe as well. Kudos to Chuck Hirstius at VMware for the work presented and have a look at his blog at www.minfluxinc.net. There are a couple of great tools for PCoIP available there.

This was my second day on VMworld in San Francisco.

He also announced that the beloved vRAM licensing model will be gone and that there will be no limitation on the CPU’s on each vSphere level. I’m really glad they are going to let vRAM go away, because I haven’t really leared it and now I don’t have to!

Steve Herrod, CTO, will hold the keynote tomorrow, so I’ll include him in tomorrow’s post.

Now for the more technical stuff. I attended a couple of advanced technical sessions regarding end user computing, basically VMware View and PCoIP.

VMware View Reference Architecture
I attended this session last year as well and last year VMware was talking about building blocks for a VMware View environment. If you want 2000 desktops, you’ll need something like this. If you need another 2000, add a block. I find the American standards of small, medium and large kind of funny. 1000 desktops are a really small environment in the US. I haven’t seen a View environment even close to that back home.
Luckily, this years’ reference architecture session was focusing more on how to solve the primary challenge with desktop virtualization; storage. The message in the presentation; “Flash, flash and more flash”. Two flash based storage “devices” was presented to solve the IOps challenges; a PCIe card from Virident and a SAN from EMC. Both the FlashMAX from Virident and Project X from EMC was able to provide disk latency below 1ms and number of IOps of 125k-225k. I have to stop by and have a look in the Solutions Exchange one of these days. Some numbers on the Project X “brick” for you;

• 1000 linked clones ready for use; 2h 25min
• 1000 linked clones recompose; 2h 6min
• 1000 full clones ready for use; 2h 45min

This is amazing when you SAN is a 3U storage array. The numbers for the FlashMAX are crazy;

• 500 linked clones ready for use; 61min
• 500 linked clones recompose: 58min

A cool thing about this is that the SAN/card was not running at 100%. This means that you can actually do deployments or recompose during opening hours without worrying about poor user experience. Isn’t that impressive? I want one of those!

VMware View Troubleshooting
VMware View installation is more or less a “next, next, finish” operation when the infrastructure is in place. It is very easy to install and very easy to manage as long as everything is working as it should. It is when it does not do what it is suppose to you’ll need troubleshooting skills.
This presentation started out talking about the new enhanced security features introduced in View 5.1. The new SSL structure will bug you with a red status until the default certificates are changed with your own. Annoying – yes. Necessary – probably.
A few links for you;
* View Persona issues; KB2019502
* PCoIP sizing; myvirtualcloud.net/?page_id=1562
* PCoIP log viewer: mindfluxinc.net

Other typical errors in View configuration:

• View composer needs access to the ESX host on TCP port 902 to do its magic.
• Black screen of Death: VMware tools have replaced the WDDM display adapter driver or firewall is blocking port 4172 used for PCoIP traffic.
• Randomly disconnects of sessions: View Agent does not successfully “check in” with the connection server.

For the first time, I decided to be here for the day 0 “lab show”. Day 0 is all about registration and labs. I went down to the Moscone Center around 9 to pick up my badge and materials – and have a look at that; A brand new VMworld backpack and drinking bottle.

I’ve chosen to give the first day number 0. The reason is simple; the only thing going on today is registration and hands on labs. This year there are a total of 36 different labs that include topics covering many of the VMware products; vSphere, vCloud Directoy, vFabric, VMware View, vCOps (vCenter Operations), Socialcast and Zimbra.

One of the new things VMware has introduced to the hands on labs this year, is the BYOD Entrance. The Bring-Your-Own-Device entrance allows you to; exactly, use your own device to do the labs. You’ll get some login credentials and a URL. After login, you get access to your lab environment using the VMware View client. Windows, Apple or Android devices can be used. I’m gonna have to use my Transformer Prime tab to enter the BYOD green room since I’m running Ubuntu on my laptop. Unfortunately, VMware does not have any supported View client for Ubuntu yet. Come on VMware! You have the client for Mac, where is it for Ubuntu?

The labs was opening at 11, so I went out for some breakfast. I should not have done that. By the time I got back, the queue was building up and had reached 2 hours. I thought; “weeell, I might stop by later”. I never did. It seems like they had some issues with the labs that hopefully is fixed for tomorrow. The Welcome reception was held at the Solutions Exchange, where all the exibitors are located. I stopped by the Teradici booth, which actually is bigger this year, and talked about new features and possibilities with the Tera 2 chip. I’ve decided to do a dedicated post on the Tera 2 later.

This is my first post in a very long time. There are several reasons, but the simple explanation; I’ve been busy. Some of the stuff I’ve been busy with will appear on this blog later on.