Novell PlateSpin Migrate 9.1 new features

Novell released version 9.1 of PlateSpin Migrate at August 5, 2o11. I have not been working with the product for a while so the release was unnoticed by me. Some new features are usefull. Mind that VMware has added some interesting new features in the free VMware Converter. A very usufull feature is the incremental synchronization. This enables to virtualize the source server in two steps. First all of the data while the applications on the source server are active. Then during off hours the incremental data. Read more here.

Release notes of Migrate 9.1 here.
New features are:

Support for SLES 9 and OES2 workloads

The addition of support for migrations of source workloads running SUSE Linux Enterprise 9 and Open Enterprise Server 2 expands the already market-leading range of workloads supported by PlateSpin Migrate.


Bandwidth throttling and compression

Bandwidth throttling gives you a greater degree of control over the amount of network resources used by PlateSpin Migrate during migration jobs. You can now reduce network strain during peak hours, and maximize network usage during off hours.

Compression also allows a greater degree of bandwidth consumption control, by reducing the amount of raw data sent over the network.


NAT Support

Support for Network Address Translation (NAT) environments allows PlateSpin Migrate to now be used in a much less intrusive way for source workloads residing behind a NAT environment. This gives you more flexibility and easier configuration for wide-area-network migrations, since PlateSpin Migrate can now be a viable migration option even in the absence of a Virtual Private Network (VPN).


Thin Disk Support

PlateSpin Migrate now has the ability to leverage the thin provisioning capabilities of the ESX hypervisor. Leveraging thin disks gives you a greater degree of control and flexibility over the total disk storage required for virtual machines. Workloads with large amounts of free space are no longer required to have the same level of free space allocated when converted to a virtual machine.

VMware Converter 5.0 incremental synchronization feature

When performing a conversion of a physical server to a virtual one (P2V)  several tools are available. VMware has been offering the free to use VMware Converter for a long time. Novell  has a commercial solution named PlateSpin Migrate.

PlateSpin Migrate used to have a unique feature which was able to create a virtual machine and transfer all the data while the source server remains online. Then, during a maintenance window applications on the source server are stopped and the changed data since the initial transfer (or delta) is tranfered (synchronization).
The synchronization feature reduces downtime of applications since only during delta transfer the application needs to be shutdown. Good for the end-users and good for the staff performing the conversion during off hours. Instead of waiting lots of hours to complete the P2V data transfer, you now can be ready in less than an hour (after post conversion tasks like removing management software etc).

Since the release of VMware Converter 5.0 the tool has a synchronization option like Migrate is offering. I used both Migrate and Converter and must say Converter works much easier. Migrate needs to prepare the target server and can have issues. Converter simply performs a P2V conversion and the admin selects to synchronize changes.  Synchronization can either be immediately or scheduled. Do not select ‘Perform final synchronization’ when you run the P2V job for for the first time.

A synchronization run can simply be started by clicking on the job used for the initial P2V and select ‘synchronize’. A click on Next and Finish and off you go. How simple is that!

As PlateSpin Migrate works with a helper VM this has much more potential issues than the vCenter Server API communication which Converter is using.
The only feature missing in Converter is the ability to automatically configure the IP-configuration of the target server. This is something which can be automated by PlateSpin Migrate. Also Migrate can be used for other hypervizors than VMware alone.  

Scheduling synchronization only works on vCenter or ESX(i) server 4.0 and above. For older ESX/VC and all Workstation/Fusion/Player only one-time immediate synchronization is supported.

Converter installs the so called “bitmap driver” which is used to track changes on the source between the last and the current synchronization. The idea is to transfer only changed sectors thus minimizing the synchronization time. Since bitmap driver once started cannot be stopped, it is set to manual start rather than an autostart. The idea is if Converter needs to track changes (synchronization) – it will start the driver on demand if not running.

The bitmap driver is installed as service on the source server. The service is called bmdrvr

 Keep in mind that scheduling and multiple incremental synchronizations have a limitation – the target needs to be a VC server or ESX version 4.0 and above. Converting to Workstation or ESX 3.5 does not allow scheduling synchronizations (however one immediate final synchronization is possible).

VMware Converter 5 Beta compared to Novell PlateSpin Migrate 9.1

Two of the most common tools used for conversions of physical servers to virtual machines on VMware platforms are VMware Converter and Novell PlateSpin Migrate. VMware Converter Standalone is free to use while PlateSpin Migrate needs a license for every workload (server to be converted). 
VMware released in the beginning of June a public beta of the next release of Converter: version 5.  
Novell is expected to release a new version of PlateSpin Migrate. Version 9.1 to be released at July 15.

VMware Converter 5 is having almost all the features that PlateSpin Migrate has. The current release of PlateSpin is missing some features which can be quite handy. Being able to thin provision virtual disk is one of those features. PlateSpin has been slow in adopting new vSphere features.  Support for distributed switches was also added quite some time after being released by VMware.

VMware Converter Standalone 5 beta.
Version 5 has two interesting new features which are currently only available in PlateSpin Migrate. The first new feature is the ability to align virtual diskfiles with the storage system. This is important to prevent performance issues. A lot can be found on Internet about alignment.
The second new feature is the ability to synchronize data after the conversion has ended. In the current version the process of a conversion is that a source physical server can be running while performing the conversion (hot clone). When the data transfer has ended a last synchronization will take place to make sure the changes in data during the file transfer are copied to the target virtual machine. This does not allow for a conversion split in two jobs in which the last job is to be run during off hours copying only the incremental data. Using PlateSpin it is possible to perform an intial creation of the virtual machine and the data during office hours. Then, at off hours a synchronization job can be performed while the applications/services are down syncing only the changed delta.

With Converter 5 there is also an option for “enhanced synchronization”, allowing to run multiple incremental updates and actually schedule them for appropriate time.

Novell PlateSpin Migrate 9.1
So far there has not been an option in the conversion job properties to set which virtual network adapter should be used. Migrate installs the flexible adapter which. This has been fixed in the 9.1 version. The default logic will be to use the best adapter available on the target host for the VM type being created.

Thin provisioning of virtual disk is a feature that finally will be available in at least PlateSpin Protect. Not sure if it will be available in the 9.1 release of Migrate.

There are still some features on the wishlist. Like  the ability to select the VMware paravirtual SCSI controller in the job (choices are now limited to LSI Logic and bus Logic). Also the ability to set the annotations to something else as the default of Migrate ’Virtual Machine created by PlateSpin on <date> and <time>’ would be nice.

The current version of Migrate can have issues with multi-homed source servers (servers with more than one network adapter). While the deployment of the agent is succesfull, the file transfer will hang at 1 % with an error saying ‘no response on port 3725. This happens because the agent/controller is using the wrong network adapter. Sometimes this can be solved by disabling all but one adapters but this workaround is not usefull for live servers.

The new version of Migrate will include NAT support, which will be beneficial in this type of scenario. Essentially, NAT support allows the user to define the IPs used for communication during the migration jobs. Being able to specify the IP on which both the PS server and source/target servers communicate has proven to be beneficial in cases like this.

Migrate  still has some unique features. It will automatically configure the IP-configuration on the target server. Using Converter first the original adapter needs to be removed before the IP configuration can be set on the virtual adapters manually. This is additional labour and can lead to errors.
Also Migrate is able to convert to Hyper-V and perform virtual to physical conversions.

VMware Converter 5 public Beta available (alignment feature)

A public Beta of VMware Converter is available for download now.

The VMware vCenter Converter Standalone 5.0 beta includes the following new functionality:

  • Preserving the LVM configuration on the source machine during Linux conversions.
  • Enhanced synchronization including options for scheduling synchronization tasks and performing multiple synchronization tasks in a conversion job.
  • Optimized disk and partition alignment and cluster size change.
  • Conversion data is encrypted between the source and the server.

 The most intersting feature is the ability to align partitions. Alignment is important to get the most out of the storage. There are many articles on Internet about alignment. See http://vmtoday.com/2010/06/storage-basics-part-vii-storage-alignment/ for example.

When Windows Server 2008 is installed on a newly created virtual machine it will align itself.

When optimized disk aligment is selected in Converter 5 (it is on by default) the partitions on the target disk are aligned on 1 MB boundary for Windows guest OS, and on 64 KB boundary for Linux P2V. If a SAN is using a different cluster size the default can be changed. On the Options page in the wizard you need to select Data To Copy page, then click on Advanced view and then select Destination tab. If volume-based cloning has been selected you have an option to change the target cluster size of each individual volume.

PlateSpin Migrate transfer speed

The time PlateSpin needs to tranfer the data from a phyiscal server to a virtual machine obviously depends on several things. The bandwidth of the network and amount of data to be transfered are the most important parameters.

Before a P2V project is started, make sure the bandwidth between the source servers and the destination servers is tested and performs as expected. I have seen some examples were there was a 100 MBps link between the source and target servers while on the same 1 GB LAN. Do not assume the network performs at the link  same speed as the link connection between server and switch. A lot of components in between source and target can influence the overal bandwidth.

It is important to get the best possible speed to reduce the transfer time. This will speed up the time needed to finish the project and more important, it will reduce the downtime of the application during the initial transfer and synchronization jobs delta and maybe even more important, the best possible bandwidth will save you a lot of extra hours during off hours.

To give an idea of the amount of time I hereby give the speed of a P2V conversion project I recently did for a customer. The speed for the data transfer was around 244 GB of data in one hour for a  block based transfer on a 1 GB LAN network .

Time from start of the job to the start of the actual file transfer is 8 minutes. Time to complete the job after the data transfer has finished is around 16 minutes.

Next release of Novell PlateSpin Migrate expected in June

The next release of PlateSpin Migrate is expected to be available in June 2011. It will have the ability to throttle the bandwidth on tranfer sessions according to this posting on the Novell site.

PlateSpin Migrate job hangs at stage 2.3.1 preparing files

While doing a ‘prepare for synchronization’ job the job hangs at stage 2.3.1 preparing files. The job is 76 % complete.

https put file failed .. Reason unable to connect to the remote server.

Solution: PlateSpin Migrate server is unable to connect to the VMware vCenter Server to upload files used for the helper VM. A reboot of the vCenter Server and restart of the job solved the problem.

VMware Converter job ‘failed at 1 %”

While performing a P2V job of a Windows server using VMware Converter the job failed at 1 %. Failed : an error occured during the conversion is displayed. The logs files are not clear in what could be the cause. Last status is “a snapshot of the source system was created succesfully.

Solution: make sure there are no network ports blocked. In this case the source server was an ISA server. After creating a rule which enables traffic between the source server and the server VMware Converter was running on, AND creating a rule which allowed traffic on ports 902 and 443 between the source server and the ESX host the conversion went fine.

For an exact overview of required ports used by Converter see this KB article

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1010056

Virtualization of an Active Directory domain controller (P2V)

If an infrastructure is going to be converted from physical servers to virtual you should consider how to handle Active Directory Domain Controllers. This posting is a reference to some articles posted on the internet.

I advise to always have at least 1 domain controller installed on a phyiscal server. This prevents the catch22 situation where the virtual infrastructure is not available and authentication is not possible because all DC’s are virtualized. For disaster recovery Microsoft DPM server needs Active Directory so that is another reason to always have at least one DC running physical.

At Microsoft TechEd 2011 North America one of the session was about virtualization of Active Directory Domain Services. See the video of the session here.

Customers are looking to further virtualize their environments: file servers, web servers, DNS servers, and even their domain controllers. It is clear that virtualization provides many benefits in areas such as deployment, disaster-recovery and lowering TCO. However, while virtualization offers many powerful capabilities and greatly simplifies repetitive tasks, it is a technology that must be handled with care when used in conjunction with Active Directory. In this session we review fundamental concepts within Active Directory and the impact of cloning and virtualization upon domain controllers, domain members and Windows in general. We also discuss how to best leverage virtualization and how to both mitigate problems and avoid occurrences in the first place.

 

There are several scenario’s thinkable to virtualize domain controllers:

1. install a new virtual machine. Install Active Directory Domain Controller role on it. Transfer roles of physical DC to the new DC and dc-promo the physical server to remove the DC role.. This is by far the best option as there is no risk for issues in Active Directory. Mind however that this involes a change of the DNS servers. So you might have to change the DNS server references of each servers, and adjust the DNS servers published by the DHCP servers. Also some applications running on the same server might be dependent of the local domain controller.

2. Demote the domain controller role on the physical server. Then perform a P2V. After that has finished, dcpromo the virtual server to a domain controller if needed. Ideal would be to create a server with no applications, just the domain controller role.

3. P2V the physical server to a virtual machine. Sometimes this needs to be done because of lack of time. Some organizations deciced to install applications on the domain controller. Manually  reinstalling the application(s) on a newly created virtual machine can cost a lot of time because documentation, media and licenses cannot be found. An exact copy of the physical server will prevent the hassle.
However, the procedure to P2V a domain controller needs some attention.
If PlateSpin Migrate is used to perform the migration, the job will ask for administrator credentials on the source server. As the active directory services on the source server are disabled (as the server is in AD recovery mode) the password check will fail! Make sure you know the Active Directory Restore Password as the account Administrator and the Restore Password is used for authentication of the PlateSpin job. 

Also you will need to have some experience and knowledge on Active Directory. There is a change the P2V -ed domain controller lost it’s connection to the domain and needs to be connected again. This could happen if the DC has been offline for too much time for example.

Read these articles for info:

How to: P2V a domain controller by Ted Steenvoorden
I performed the procedure described above several times and had no problems. Make sure all FSMO roles are moved from the server which needs to be P2V-ed to another domain controller as these roles are not available when the server is in Directory Services Restore Mode .  Also check Global Catalog server role.  

Virtualizing a domain controller, how hard can it be? by Gabrie van Zanten   

P2V a Domain Controller? Why would you? by Christian Mohn

Converting domain controllers by Duncan Epping

Virtualizing existing domain controllers by VMware

PlateSpin Migrate data transfer troubleshooting

One of the most common problems on using PlateSpin Migrate is the data transfer process. After the target VM has been created and booted with Windows PE, a process is started on the target to receive files from the source server. The process which does the transfer is ofxcontroller.exe . The data is send from source to target over port 3725 TCP. There can be a number of reasons why the data transfers does not get further than 1% and stalls / hangs . If the process is at 2% most of the times the data tranfers will succeed to 100%. After a succesful data tranfer, the job almost always complete with success.  

Reason for stalling:

- firewall is enabled on source server
-firewall is blocking traffic between source and target
-issues with OFX controller on server
-OFX controller is using wrong network interface on source 

If the data transfer cannot be started, errors will appear in the job log

no response on port from host <ip-address>
no response from receiver on port 3725
the file transfer is stalled
unable to connect to port 3725 on ip address <ip address of source>

Most common reason for these issues are firewall related. Make sure there is no firewall between the source and target blocking traffic on port 3725. Make sure the firewall on the source server is not blocking port 3725. If so, add a rule to allow traffic or disable the firewall if possible.

Perform a ‘netstat -ano’ command on both the source and target server to verify a process is listening on port 3725. The target Windows PE does not have netstat command available. Perform a ‘net use ‘ command to map a drive on a server which has it available and perform netstat from the mapped drive. Search in the ‘netstat -ano’ output for a line like this

TCP    0.0.0.0:3725           0.0.0.0:0              LISTENING       260

I had an issue on virtualizing a Windows 2003 server acting as a proxy. It had an internal IP address on the internal network interface  and several external ip-addresses on the external interface. Each time I tried to perform a P2V job, the job stalled at the data transfer stage. In the job log it listed the controller on the target was not able to connect to the ip address on the external network interface. I wanted to have the controller connect to the internal interface. Even after adjusting the properties of the source server and deleting all external ip’s in the properties, the log file kept reporting the external ip address.
PlateSpin support explained that the controller will try to connect to all IP addresses which are available at the source. I do not think this is the case!

We had a server with two nics. One nice was configuered with an IP-address, the other not and was disabled. When we did the P2V the job stalled at 1 %, saying port 3725 was blocked. What we did was changed the IP-configuration to the second nic, disabled the  first nic and created a new job. Thereafter the job performed well !

A reboot of the source server might help as well.

To troubleshoot connections, a telnet to the source server on port 3725 can be performed. To do this, for VMware targets, start VMware vSpere client, connect to the console of the target VM. In the Windows PE console, connect to the source server by net use y: <a href="//\\\\<source ip>\c$ . Locate telnet.exe and perform a command: telnet <source ip address> 3725 

As soon as  the job is at stage 6.8: Copying Volume Data from Source to Target started (1% progression). the telnet command should result in a connection and you probably will see strange characters. That is good. If the response is something like ’not connected’ port 3725 is not reachable and a firewall is likely blocking traffic.

Novell has a knowledgebase article available describing the issue. The solution involves installing a Microsoft hotfix. The article can be found here.
My experience is that in most cases the problem is related to firewalls or source servers with multiple nics.  

If some other process is using port 3725, it can be changed by the following commands:

1) Open “C:\Program Files\PlateSpin Migrate Server\Web” directory.
2) Open the productinternal.config file in a text editor.
3) Locate the following line under <optimizedFileTransferSettings> and <legacyFileTransferSettings>:

<fileTransferPort>3725</fileTransferPort>

  4) Change the 3725 value under both <optimizedFileTransferSettings> and <legacyFileTransferSettings> to the desired port number and save the changes.

HP Virtual Connect for Dummies second edition released

HP released a second edition of the popular HP Virtual Connect for Dummies PDF book. The second edition has information on FLex-10 which the first edition did not have.

Download directly from this link : HP_Virtual_Connect_for_Dummies_2nd_edition

PlateSpin Migrate stalls at stage 3.1.1 Creating virtual machine on ESX server (running 1% complete)

While performing a P2V job using PlateSpin Migrate, the conversion job hangs / stalls at stage 3.1.1 Creating virtual machine on ESX server (running 1% complete).

This could be caused by:

-the ESX hosts on which the target VM is hosted is a running the free ESXi edition. The API’s which are used by Platespin to create a virtual machine are not available on the free edition. The evaluation version will work and all licensed editions will obviously work as well.

-the VMware datastore is not able to store the VMDK of the target VM. If the datastore has a blocksize of 1MB, a VMDK or other file can be  256GB in size max. If a partition on the source server is larger, it will not work.

In my case, both were okay. So I deleted PlateSpin Migrate server and client, deleted the SQL Express database, deleted IIS role , deleted the Program Files folders or Platespin and SQL and rebooted.

After reboot, I disabled User Access Control and reinstalled PlateSpin. After that, the issue was gone!

It might be caused bu UAC being enabled! Platespin advises to disable UAC.

PlateSpin Migrate discovering server details hangs at 1%

When discovering source Windows servers, the job hangs at 1% in ‘discovering server details’.

From the PlateSpin Migrate server, the source server can be pinged. But from the source server the PlateSpin server cannot be pinged. IP configuration is okay.

Make sure the firewall running on the PlateSpin server allows traffic used by PlateSpin. Quick and dirty solution is to disable the firewall. You might need to reboot the server to completely get rid of the interference of the firewall.

New versions of Novell PlateSpin Forge, Migrate and Protect released

At October 4 Novell released new versions of PlateSpin Protect (version 10.0.2), PlateSpin Forge (3.0.2) and PlateSpin Migrate (9.0.2)

Download here:
http://download.novell.com/index.jsp?product_id=&search=Search&families=&date_range=&date_start=07+Oct+2010&date_end=07+Oct+2010&keywords=platespin&sort_by=&results_per_page=&x=0&y=0

PlateSpin Migrate 9 review

Novell released PlateSpin Migrate version 9 end of June 2010. This is the first version to support distributed switches available since the release of VMware  vSphere. Below are my experiences so far with this new version.

The release notes can be found here

The upgrade from version Portability Suite 8.1.3 to version 9 went without any problem. The upgrade took around 30 minutes. Before performing the upgrade, I de-installed the Portability Suite Client as described in the Install Notes.

The client has some minor changes in the user interface. I noticed it is still not possible to run two instances of the client on the same server. A pity as this is very usefull during conversions when several staff members are performing conversions at the same time. The alternative is to install the client on a workstation or other server.

I noticed the default job values were reset. I prefered to have the display name of the virtual servers without the ‘_VM’  string attached by default by PlateSpin. After the upgrade the _VM suffix was back.

Also , Migrate 9 still creates at random choosen datastores a folder called Platespin. This folder holds several files used for the conversion process. One of the files is a floppy image.

PlateSpin Migrate immediately displayed the  portgroups which are created on distributed switches of ESX4 servers. I did not have to perform another inventory of the vCenter Server 4 nor the ESXi4 hosts even I discovered those using Migrate version 8.1.3. This is nice.

The conversion speed does not seem to have changed. We still see an average data transfer speed during P2V’s filebased transfer of 40 GB per hour.

Platespin Migrate is lacking an important feature if you decided to use thin provisioned disks. Platespin migrate 9 does not support thin provisioned disks. So if you want to convert your physical server to a virtual, Platespin will create thick provisioned disks. VMware Converter however is perfectly able to create thin thick during the P2V process.
To solve this using Platespin, after the P2V is done, you will need to storage migrate the discs to another datastore, and storage migrate back to the original datastore selecting thin provsioned discs. A time consuming process.

Unfortunately, the ability to create thin provisioned discs is not on the roadmap for future releases at this moment. (september 2010).

We did a couple of P2V conversions of servers having a lot of data. Over 1.5 TB of data in many small files. Because the partitions of the physical server were too small, we needed to enlarge the target volume. This means instead of the faster blocklevel copy, we had to select the filebased transfer. We had several jobs that aborted or seem to hang at 1% copying files. This is an issue which is described in the knowledgbase of PlateSpin. Solution is to select the blocklevel copy.

Also we had an issue while performing a P2V conversion to either ESX4 or ESX4.1 hosts. We are using vCenter Server to discover hosts and virtual machines.

Platespin 9 is a must have if you are performing lots pf  X2X conversions to vSphere 4.x platforms and using distributed switches for your vm portgroups, when minimum downtime is a must. However, prepare for various issues which can be caused by a lot of things. PlateSpin is not an easy to use tool which guarantees succesfull conversions.

Follow

Get every new post delivered to your Inbox.

Join 75 other followers

%d bloggers like this: