Modifying VMware Multipathing Policies with EMC’s VSI Path Management Plugin

Here is a handy tool for setting and modifying multipathing policies in VMware when using EMC backend storage.  It supports CLARiiON, VNX, Symm, and VPLEX devices. It also supports both PowerPath and NMP.  The plugin is free to EMC customers with a PowerLink account.  To download, login to PowerLink and navigate to: Home > Support > Software Downloads and Licensing > Downloads T-Z > Virtual Storage Integrator (VSI).

Configure SRM 4 with SQL Express 2005

I found this document very useful when setting up SRM 4 in the lab recently.  The part that really caught my attention was that the sql account needs to have at least “bulkadmin” privileges.  Of course, “sysadmin” will work as well. 

http://communities.vmware.com/docs/DOC-11547

Finding Drive Speeds for CLARiiON Disks Using Powerlink and Navi/Unisphere

Ok, so it’s common knowledge that drive speeds are shown on the front of all EMC disks but what if you don’t have physical access to the array?  I often find myself in this situation and the ability to lookup drive speeds without access to the array has been very helpful.  This post will go over the process for locating the part number for the disk and cross referencing it with the Disk and FLARE OE Matrix.

1. Find the CLARiiON TLA Part Number for the disk or disks which drive speed needs to be determined. This can be done a couple of different ways.  If there are only 1 or 2 disks which are in question, then it will be quicker just to right-click on that drive in Unisphere and bring up the properties.  If speeds of multiple drives are in question then generating a physical disk report would be a better approach.  I give examples of both methods below.
    
    
2. Now that we know our part numbers, we can cross reference them with the appropriate disk matrix which can be found on Powerlink.  If we look at emc253342, it tells us that:

 On Powerlink the Disk and FLARE OE Matrix documents can be found under the model specific Installation/Configuration section in the Technical Documentation and Advisories area of Powerlink.

For example:  If you are looking for the Disk and FLARE OE Matrix for a CX4-960 model array, Navigate to: Support > Technical Documentation and Advisories > Hardware/Platforms Documentation > CLARiiON CX4 Series Systems > CLARiiON CX4-960 > Installation/Configuration

Note: The Disk and FLARE OE Matrix documents are based on the product series and not individual array types. This means the CX4 series that includes the CX4-120, CX4-240, CX4-480, and the CX4-960 will all be contained in the same document.

So by searching this matrix we finally find our drive speed. 

FAST and FAST Cache: Some Quick Highlights

Edit: A quick response on twitter pointed me to a more current config chart which is now updated.  The newer chart is referenced at emc251589

Have you just upgraded your CX4 to FLARE 30? Are you super excited about using FAST but just don’t know where to start? It’s an exciting time for sure in the EMC storage world and I wanted to share some highlights I’ve gathered around using FAST and FAST Cache on your Unified Storage Array.

First of all, a note about FAST sub-LUN tiering.  FAST is cool but let’s keep good perspective on IOPS.  From my understanding, FAST does not dedicate IOPs to any LUN or application.  For many applications (I would dare to say most applications), sharing IOPs in a FAST environment will have little to any impact if sized appropriately with the right number of Flash, FC, and SATA Drives.  However, if there is an application that needs dedicated IOPS, by all means use traditional RAID Groups.  I’ve heard it mentioned that Storage Pools are implemented more from a Storage Admin perspective where automation and ease-of-use are sought after.  On the other hand, RAID Groups are implemented more from a Storage Architect perspective where required performance metrics are known and IOPS can be dedicated.

FAST Cache

• Flash drives do not need dedicated DAEs or back-end buses.

• Flash drives are configured as RAID 1 mirrors

• Can be applied to any LUN
• FAST Cache is disabled by default on a LUN created on Flash Drives ( I can’t imagine why FAST Cache would be needed on a LUN that already lives on SSD)

FAST

• Utilizes heterogeneous pools to allow combining multiple storage devices
• Does not differentiate between different drive speeds.  For example, FAST does not know the difference between a 15k FC drive and a 10k FC drive.
• It is recommended that drives with different rotational speeds are kept in separate storage pools.  This allows for consistent performance between storage tiers in the same pool.
• Performance is analyzed every hour and all data within a FAST pool is given a rank
• Auto-Tiering is based on a schedule and should be kicked off after hours to maintain minimal performance impact.
• Manual relocation is available if needed.
• Auto-Tiering is configured at the array level.  Manual relocation is performed at the storage pool level.

FAST and FASH Cache Comparisons and Recommendations

• When using Flash drives and combining both FAST and FAST Cache, consider dedicating the Flash drives for FAST Cache and only include FC and SATA drives in FAST pools.
  • FAST Cache is array wide and will give immediate performance benefit to any I/O intensive bursts of data.  FAST sub-LUN tiering will move warmer data to FC drives and colder data to SATA drives.

• FAST Cache operates using 64 KB chunks while FAST operates using 1GB  chunks of data.

Below is a comparison chart from this EMC Whitepapaer

Another important matrix to keep in mind is the configuration options.  This version if from emc251589.

Citrix Streaming Issue trying to call Microsoft Office causes long delay

Looks like there is a bug using version 6 of the offline plugin causing a 5 minute delay trying to open Microsoft Office from a streamed application.  The issue I speak of presented itself when trying to call Microsoft Word (Either locally or streamed) from within a streamed clinical application.  The clinical application I was working with attempts to open Word to generate a from using Word macros.  I tested the issue with Word installed locally on the XenApp server as well as having Word installed in the streaming profile.  In both scenarios I encountered the 5 minute delay when trying to open Word from within the streamed app. 

I contacted Citrix support to further investigate the problem and was surprised to find that this was an issue known within the Citrix support team, however, has not been released as a KB on support.citrix.com.  The fix is relatively straight forward and only involved changing the HooksMSI DWORD value in the registry to 0.

HKLM\SOFTWARE\Citrix\StreamingHook\AppInit_DLLs\CtxSbxHook

Bug in Citrix XenApp HR06

I was trying to lauch a published app in a client’s new XenApp 5 HR06 environment and kept encountering the following error:

The application failed to initialize properly (0x06d007e). Click on OK to terminate the application.

Launching the app locally and via RDP worked great so we knew it had to be something from our ICA session. A quick google search by the client turned up this forum article: http://forums.citrix.com/thread.jspa?messageID=1453809&

We were able to resolve the issue by replacing the “C:\Program Files\Citrix\System32\icammdrv.dll” file with a previous version copied over from a HR04 version.  Sure hope Citrix comes out with a hotfix for this soon.

Implementing RecoverPoint via Deployment Manager

I just finished an implementation and management course for RecoverPoint this past week.  Something that caught my interest from the very beginning was how the setup phase of the  course was modeled around the new Deployment Manager.  As the tool has evolved, this now seems to be the de facto method for installations, upgrades, replacement of existing appliances, and the addition of new appliances.  So, the big question… Does it work?? Well, for the most part yes, it works quite well.  I have seen a few issues when using the RecoverPoint/SE option but for the other setup options the tool appears to work very well. So, even if you’re doing an SE install be sure to choose the RecoverPoint option not RecoverPoint/SE. (This is for version 1.0.  Future versions may offer better outcomes using the SE option)

Some Prereqs for the Tool

• You will need to have the same version of code on all of the RPAs
• All Fibre Channel connections and zoning need to be setup properly
• Each appliance needs to see the Repository, Journal, and Replication volumes on the same site
• You need access to an RPA from your installation station
• SSH needs to be open from your installation station to one of your RPAs

The Process

The install wizard pretty much walks you through the rest. You will enter your configuration options and  the wizard will verify connectivity to all RPAs and attempt to detect any errors with the existing configuration.

After all information has been entered throughout the wizard, the deployment tool will begin to install you clusters and configure your sites.  There isn’t a lot of helpful information given around the status and progress of the install.  If a continuous ping is issued to all of the RPAs, momentary lapses can be seen in connectivity as the RPAs reboot as part of the process. That’s really all there is to it.  After the install is complete, try connecting to the site management IP and launch the RecoverPoint management tool.  If all is well then you’ve successfully implemented a new RecoverPoint install from the GUI using a straight forward single interface.

Run this script after modifying XenApp Prep images

If you’re using XenApp prep and working with Provisioning Server or virtual machine templates, it’s a good idea to delete the service and rerun the initialization command when modifying your image.  I usually create a batch script that can be executed before shutting down the server after all changes are made.  The commands to execute are:

net stop xenappprep
sc delete xenappprep
xenappprep.exe /PVS

Note that the /PVS switch is only used if working with Provisioning server

Changing ODBC Connections to Site Recovery Manager Database

I recently encountered a situation where I needed to change the ODBC connection to an SRM database.  The ODBC had originally been setup to use Windows Authentication and I wanted to use SQL Authentication.  I originally attempted to modify the existing DSN which was unsuccessful.  This error kept appearing in the SRM Logs.

[2010-03-08 13:40:07.845 04764 error 'Vdb'] Connection: Could not connect to database: -1

After scratching my head and doing some research, I found some good information on solving my dilemma.  The first step was to create a new DSN using SQL Authentication and then to update vmware-dr.xml with the name of the new DSN. Next, I followed some commands outlined in this KB:  http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1008426

I first ran:

installcreds.exe db:new_SRM_DSN Administrator

Replace “Administrator” with the username of the SQL account specified in your new DSN.  It seems that in SRM 4.0 the syntax of the command has changed slightly to look like this.

installcreds.exe -key "db:new_SRM_DSN" -u Administrator

After this completed successfully, I had to modify my connection to Virtual Center again using the srm-config command.  My colleague, Jeremy Waldrop, has posted a nice blog on running this command and you can check it out here:  http://jeremywaldrop.wordpress.com/2009/05/01/vmware-site-recovery-manager-service-account/

You should now be able to connect with your new System DSN.

VMware Partner Exchange 2010

I’m a little late on this post I know, anyhow, better late than never right?  VMware PEX 2010 was two weeks ago and a lot of information around its contents has already been discussed and talked about. However, I do want to share my experience around the event and my take on things that were discussed. I know some material is NDA and I’m doing my best not to violate any of those agreements in this post.

ESX Classic vs. ESXi

There’s been a lot of buzz around the transition to ESXi.  To keep it simple, ESXi is VMware’s preferred hypervisor and partners and customers alike should start looking at making the transition soon.  It appears that in the very near future, ESXi will be the only hypervisor offered and VMware will announce an end of support date for ESX Classic.

Vblocks

I have worked with EMC products for years and I signed up for the EMC bootcamp at PEX. To start of the session, Chad Sakac discussed VCE and Vblocks.  Vblocks are cool.  There’s so much effort going into the VMware/Cisco/EMC alliance and I was really excited to see the opportunity that a Vblock will present.  A few reasons I would recommend these to my customers

• One Bill of Sale – No buying several different parts and pieces from different vendors.  This should greatly simplify the planning and purchasing phase of the project.
• One portal for support – No more juggling 3+ vendors for support of your Virtualizatoin environment.  Whether it’s a storage, server, networking, or VMware problem there’s a single portal you can access for support.
• Unified Management Interface – EMC is developing a single interface to manager your environment using Ionix

Much More Cloud Talk

The cloud was the main topic of almost every event.  All of VMware’s resources seem to be focusing on making public and private clouds easy for everyone.  A lot of effort is going into “Project Redwood” which allows customers to access resources On TAP.

Virtual Desktops

Virtual Desktops are a huge part of VMware’s road map.  Future versions of View will be easier to manage but very powerful.

Technical Content

I was really impressed with the technical content offered at PEX.  I’m not going to go into detail in this post on all of the technical sessions but I did want to give VMware props on putting together some great content.  For any engineers debating whether to attend PEX in the future, I would recommend it.  There was a plethora of good technical resources.  This was my first PEX and I was skeptical before attending, thinking that the content would be mainly pre-sales focused.  This isn’t the case and it was a great environment for pre and post sales alike.