VMware Communities : All Content - Enterprise Strategy & Planning

Selecting a SAN (Interpreting Perfmon Disk IO Counters)

rgm34 — Fri, 27 Nov 2009 02:42:28 GMT

I have eight servers in my LAN Room, four are Windows 2003, the others are Windows XP boxes that don't do a lot. I also have an MSA 1000. I am looking to upgrade my SAN and implement VMware in 2010. Replication to a DR site is a future requirement.

A number of vendors have been in for demos and all of them have asked me to run Perfmon collecting the counters listed below for a 24 hour period and email them the files. None of them will tell me what they are looking for. I can open them in excel but beyond that I'm not sure what they are doing with the numbers. People in our office travel often so I've been collecting these for two weeks and really don't want to ftp all of the files to the vendors.

\PhysicalDisk(_Total)\Disk Reads/sec
\PhysicalDisk(_Total)\Disk Writes/sec
\PhysicalDisk(_Total)\Disk Read Bytes/sec
\PhysicalDisk(_Total)\Disk Write Bytes/sec
\PhysicalDisk(_Total)\Avg. Disk Bytes/Read
\PhysicalDisk(_Total)\Avg. Disk Bytes/Write

Can someone tell me what I should be looking at, I want to make sure I purchase the best system for our environment.

Are there other Perfmon counters I should be collecting?

Thank you.

Robert

Large Disk Limitations and the 2TB barrier - What problems will I face in the future

vSeanClark — Thu, 26 Nov 2009 15:01:51 GMT

With dynamic disks you might have to be careful because not all backup software providers can handle dynamic disks properly.

What's wrong with using guest iSCSI initiator to access this storage? Although there are some cons, I think the benefits of easier management, SAN tools to manage, etc., are worth it.

$.02

If you value your karma, please consider awarding points for "Correct" or "Helpful".

Sean Clark - vExpert, VCP - http://twitter.com/vseanclark - http://seanclark.us

VMware ESX VI3 and Meditech c/s 6.0 Implementation

vSeanClark — Thu, 26 Nov 2009 14:44:55 GMT

I don't have specific info on 6.0, but I virtualized Meditech c/s 5 background servers in March of 2006. It was a resounding success with only one problem. We should have ignored their hardware specs and only do a single vCPU VM for background servers. We did 2 vCPUs according to their instructions and it was a wast of system resources and actually slowed the environment down a little bit. My understudy changed these over to single vCPUs a couple years later.

I also understand that he has since dipped his toe in the water with virtualizing file servers as well. Biggest advice there is to communicate in advance what you're doing with Bridgehead and get them to help advice on correct SAN integration software and scripts needed. They will give you crap about going rogue, but they and Meditech will eventually help you.

Holler if anyone has questions. I love helping folks virtualize their Meditech systems and sticking it to the high dollar and low value Meditech system integrators.

If you value your karma, please consider awarding points for "Correct" or "Helpful".

Sean Clark - vExpert, VCP - http://twitter.com/vseanclark - http://seanclark.us

Intel processor purchase advice

meistermn — Wed, 25 Nov 2009 18:15:08 GMT

2 Socket Servers (only Intel)
For 1 and 2 vcpu VM i put them on intel nehalem ep cpu servers. e.g. HP DL 380 G6

4 socket Servers (only AMD)
For 4 and 6 vcpu VM i put them on AMD Instanbul servers e.g. HP DL 585 G6

HP HP ProLiant DL585 G6
VMware ESX v4.0 VMmark v1.1.1
29.95 @ 20 tiles
View Disclosure 4 sockets
24 total cores
24 total threads 07/14
http://www.vmware.com/files/pdf/vmmark/VMmark-HP-2009-07-14-dl585g6.pdf

I do not invest in old server architecture from Intel 4 socket X74XX CPU (Like HP DL 580 G5)

I only by 4 GB Dimms

8 and 12 or vpcu VM's i would place on AMD CPU'S on March 2010 or Intel's Nehalem EX CPU.
Look at the AMD server plattform presentation
http://phx.corporate-ir.net/phoenix.zhtml?c=74093&p=irol-analystday

or intel nehalem pdf
http://download.intel.com/pressroom/pdf/nehalem-ex.pdf

Cisco Switch Port Config

VMFIZZ — Wed, 25 Nov 2009 13:27:27 GMT

Hello all, I'm setting up my ESX server for the first time and would like to know what the best configuration for my switch ports will be.

I have two 3750s with gigabit ports and on my ESX server My production vSwitch has three Gig nics and the vSwitch for IP Storage (ISCSI) also has three Gig nics.

At the moment the switch ports are configured like this.

switchport access vlan 99 (Server Vlan)
switchport mode access
switchport nonegotiate
duplex full
speed 1000
spanning-tree portfast
spanning-tree bpduguard enable

Thanks in advance

pvscsi change on NFS - worth of doing?

ablej — Tue, 24 Nov 2009 19:35:29 GMT

I would think if the VM doesn't need it than don't do it. If the VM is only at 20% utilization it won't see the benifits of the PVSCSI. It's probably best if PVSCSI is used on a case by case basis, such as for mail, database, etc. applications.

David Strebel
www.holy-vm.com

If you find this information useful, please award points for "correct" or "helpful"

New !! Open unofficial storage performance thread

JonKnaggs — Tue, 24 Nov 2009 01:11:03 GMT

Hi, we have a pretty basic setup, and probably far from best practice. But here are our results. Our array takes a massive hit on 60% random 65% read. Looks like our results are pretty average. I think I should look at seperating the iSCSI traffic, at least onto its own VLAN. I wont be able to get a dedicated switch.

VM tested is on our busiest ESX host and LUN.
Antivirus was on during test.

ESXi 4
MD3000i Dual Controller w/146GB 15K SAS
2 x Dell 2950 (32GB, 2 x E5450@3.00GHz (8CPU), 2 x Broadcom Embedded, 1 Intel Quad GBe Card)
1 x Cisco 3560G for all traffic. iSCSI is not separated onto its own VLAN unfortunately.
2 x Gb nic for iSCSI traffic using software initiator and 2 x Gb nic for VM network traffic on each host.

VM: Windows 2003 Server R2 (32Bit)
7 Disk - Raid 5
No Jumbo Frames
-----------------------------------------------------------IOPs-----------MB/s--------RT---------CPU
VM1: Max Throughput-100%Read-----------3292.766-----102.899----18.148----7.201
VM1: RealLife-60%Rand-65%Read---------845.845-------6.608--------56.374----10.01
VM1: Max Throughput-50%Read-------------2572.745-----80.398-------22.614----7.232

Sorry for formatting.

Second storage option

Collin09 — Mon, 23 Nov 2009 21:53:00 GMT

We are about to move on to the next step of buying a storage for the VM server but I need to know what will be an option for fault tolerance.
We are thinking of getting the Dell Equallogic PS4000e for a storage space for the 2 VM Host server for HA.My concern is hardware failure can happen anytime and I would like to set up another alternative solution to recovery while the primary storage is under repair.

The ideal setup will be buying a second PS4000e unit but that will be over my budget. What will be another option that we can setup for a minimal budget for a backup for this kind of setup.

I am thinking of buying the MD3000i (next year budget) for a failover storage or some kind of replication process.If something happen to SAN (PS4000e)then I can switch to the other SAN (MD3000i).

Your input is greatly appreaciated.

Thank you.
Collin M.

How Many NICs for the Service Console

VMFIZZ — Mon, 23 Nov 2009 20:18:20 GMT

Wow thanks for the fast response! That's sort of what I was thinking, but being new, I wasn't sure how important redudancy is with the Service console. I like your resonse I will do that. Thanks again.

Good View Test Environment ... Openfiler etc

s1xth — Mon, 23 Nov 2009 17:31:56 GMT

I would like to set up a View test environment to demo the new features and to show to some executives at my company. What would be a basic and decent View test environment to show this? I am thinking the following would be good since I have used it in the past..

Openfiler iSCSi setup (or any iSCSI based program)

vCenter (obviously)

2x old PE860 Intel Xeon servers

Any other ideas or better test setups that I could use or anything I should be aware of before setting it up?

Thanks!

Training Documents for ESX4 & VC4?

Texiwill — Mon, 23 Nov 2009 16:36:44 GMT

Hello,

Most if not all 'training materials' out there at the moment are copyrighted so we cannot just hand them over. You may wish to team up with someone who does own the material to present said training, or create your own.... Outside of courses from VMware and two or three other providers there is not much publically available.

Best regards,
Edward L. Haletky VMware Communities User Moderator, VMware vExpert 2009
Virtualization Practice Analyst
Now Available: 'VMware vSphere(TM) and Virtual Infrastructure Security'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|
Virtualization Security Round Table Podcast

Boom! Fusionio + Datacore/Falconstore the new killer storage!

meistermn — Mon, 23 Nov 2009 15:31:04 GMT

And here something to play for esxi on a whitebox and a fusioio card.

http://www.amazon.com/s/qid=1258151243/ref=sr_nr_seeall_1?ie=UTF8&rs=&keywords=ioxtreme%2080&rh=i%3Aaps%2Ck%3Aioxtreme%2080%2Ci%3Aelectronics

Health Check Analyzer Appliance---Where did they hide it?

HamR — Mon, 23 Nov 2009 05:50:41 GMT

As of Nov 23, I found it by clicking on "Content" from the top link ribbon then searching for "health".

The search option on the left-hand window of the Partner Central homepage returns nothing for the same search string.

HamR.

SAN Storage physical Movement Plan with out downtime.

meistermn — Sat, 21 Nov 2009 17:16:47 GMT

Here are my experience from my storage migration:
I migrated 595 VM's and 35 TB from an old netapp 6070 storage to a new netapp 6080 storage .
In the frist step all esx host were on esx version 3.5. The VC was version 4.0.
After I migrate some VM's per storage VMOTION, i got a time out at 8 % and the failture message was "Source detected that destination failed to resume.",.
So looked at vmware http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1010045
For me this was not the option, because I don't wanted to reboot 595 Vm's.
If you can , maybe this powershell script http://ict-freak.nl/2009/11/17/vsphere-storage-vmotion-fails-with-a-timeout/ help's .

After that I decide to upgrade the esx version from 3.5 to 4.0. And bingo, I had no storage vmotion timeouts anymore!!!
ALthough this helped a lot to fasten the upgrade from esx 3.5 to 4.0.

Result was : 22 ESX Hosts were upgrade . Took me 3 days.

585 VM's were storage vmotioned without downtime . The esx hosts are HP DL 585 G2 and HP DL 585 G5
10 VM's were storage vmotioned with downtime, because the esx hosts HP DL 585 G1 cannot be installed with esx 4.0.

Next Step was to automated the storage vmotion from a csv file and vc schedule tasks with a powershell script:
Look at at svmotion script
http://communities.vmware.com/message/1405978#1405978

The total migration took 1 Month ! That was great.

Lessions learned: ESX 4.0 uses storage vmotion version 2 . esx 3.5 usses storage vmotion 1.

Other infos:
http://virtualgeek.typepad.com/virtual_geek/2008/12/real-world-experiences-using-storage-vmotion.html
Vbscript storage vmotion:
http://vikashkumarroy.blogspot.com/2009/10/storagevmotiondoc.html
Powershell:
http://blogs.vmware.com/vipowershell/2008/06/fun-with-powers.html

Suggestions for end to end monitoring solutions

aenagy — Sat, 21 Nov 2009 01:03:19 GMT

If you are still looking for suggestions, then EMC SMARTS and IBM Tivoli Monitoring. Our company did a PoC on both back in 2007 and ended up picking the IBM solution, in part because EMC had just released version 6.0 and the VMware integration was not there yet (as I recall), and the IBM solution was more flexible. I should point out that the ITM solution we evaluated was not the traditional TMF (Tivoli Framework), but the new Netcool Omnibus products. EMC had a presentation on enterprise management for vSphere at VMworld 2009 which turned out to be a pitch for SMARTS, which has been renamed to Ionix.

We were looking for an end-to-end physical and virtual monitoring/alerting/correlation solution. This meant integration with x86 (Windows and vCenter/ESX, iSeries, mainframe, storage (EMC), network (Cisco) all the way from the physical layer up to the application layer. Correlation of various pieces at the various layers was important. One question I asked each vendor during thier presentation was: is it possible to isolate performance problems of a virtual machine (in this case VMware Infrastructure 3) with a particular host and Datastore, all the way back to a specific LUN and/or HDD? Both answered in the afirmative.

Even though we have standardized on ITM for the enterprise monitoring solution, that is not preventing us from evaluating other monitoring solutions for our vSphere project, just as long as the datastream ends up in ITM. We are therefore evaluating: (1) ITM VMware VI agent (the ITM infrastructure has already been deployed) which would give us one-hop monitoring, (2) Quest Vizioncore vFoglight, and (2) Veeam nworks as we also have SCOM 2007 with the Microsoft/Engyro connector to TEC (Tivoli Enterprise Console).

In all cases I am curious to see if ITM can provide all that IBM has promised.

IBM DS3400, NetApp FAS2050, or Equillogic PS6000XV for Production Environment

djaquays — Fri, 20 Nov 2009 20:52:14 GMT

We currently have a NetApp FAS2050 (2, in fact) one that runs our production VMs (NFS) and another that is our offsite DR (snapmirror) and backup (snapvault) target. We have 5 ESX hosts and about 40 VMs that are either SLE, Windows Server or NetWare. If our CPU usage averages 10%, it's been a crazy day. Normally the average sits around 3-5%. That being said.. if I had the option to go back and do things over again.. I would have waited a couple more months for the 2040 to come out. FAR more flexible and allows you to use the 24 drive SAS trays instead of the 14 drive FC trays without sacrificing connectivity to iscsi/nfs clients (by using the single PCIx slot).

Possible Reuse of a Cluster?

Texiwill — Fri, 20 Nov 2009 17:06:39 GMT

Hello,

VMware View is one option. If you want to go a less expensive route you could use VMware Server or ESXi on the systems. Create the 10 VMs and then just RDP in to these systems using a secure VPN technology such as OpenVPN....

For large number of target VMs (which differs for everyone) I would go the View route or perhaps XenDesktop.

Best regards,
Edward L. Haletky VMware Communities User Moderator, VMware vExpert 2009
Virtualization Practice Analyst
Now Available: 'VMware vSphere(TM) and Virtual Infrastructure Security'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|
Virtualization Security Round Table Podcast

iSCSI, Linux I/O scheduler, VMFS

danpritts — Fri, 20 Nov 2009 16:14:29 GMT

Red Hat suggests use of "noop"

http://kbase.redhat.com/faq/docs/DOC-5428

This leaves i/o scheduling to ESX and/or your storage. Your guest knows very little about the actual disk layout; it's many layers of virtualization deep. Makes little sense to have your guest try to optimize in this situation.

It probably won't hurt anything, but it might; and it is a waste of time for the guest to do things that vmware or the storage controller is going to reorder anyway.

NetApp Metrocluster: "disappointed"

king@it.ibm.com — Fri, 20 Nov 2009 13:03:32 GMT

if we look into the metrocluster requirement in term of bandwidth and network configuration, you could also consider to turn your " DR " to be an hot site

That's what a Metrocluster is indeed. I would not see it as a "special/closer DR"... I would define it as a "stretched single cluster".

you may consider to have 2 hot site and cluster each others

Again .. that's the idea of metrocluster / Campus type of designs.

< are you sure you will get the person in charge on time to make a call?

I am... otherwise he/she is in the wrong role..... (if he/she is away for an extended period of time he/she will have deputies etc etc). A real DR execution may require multiple discussions and multiple level of commitments. When I was working with a big bank they told me they may decide to stay off-line for a few days to fix the issue and not commit to the DR process..... for them it would be such a big thing that they are thinking to leave the production at the DR site indefinitely .... (i.e. no failback).

This is one of the big differences between Campus and DR.

Massimo.

vConvert on an EFI based I7 IBM Server not working

LEPA — Fri, 20 Nov 2009 11:56:21 GMT

Same problem here. Trying to convert Windows 2008 Server SP2 into a virtual machine using vCenter Converter 4...
Also on a IBM 3550 M2 server.

Any help would really be appreciated!
Thank you

Need Help - Performance & Tuning

HALOTEQ — Thu, 19 Nov 2009 19:41:18 GMT

Thanks Troy!

VERY HELPFUL!!!

Q's on setting up new ESX infrastructure and equallogic SAN

huwy — Thu, 19 Nov 2009 15:53:47 GMT

Cheers. I've setup jumbo frames on the switches (dell powerconnect 5424s).
I haven't explicitly setup jumbo frames on the equallogic but if you look at the attachment it says the NICs on the equallogic box are operating at MTU 9000.
So I guess all is good.

-H

Best Practices for Virtural Center Access?

Texiwill — Wed, 18 Nov 2009 18:03:52 GMT

Hello,

The customer Role applied the Folder (VM and Template View) for the VMs you want those support admins to reboot is the best way to do it. They can then reboot via the web interface.

Even better would be to use something like Hyper9's VMM so you can reboot from an iPhone, etc.... Also this interface could provide a proxied service for the other people so you do not need to open up the entire VC or install vSphere client anywhere.

Best regards,
Edward L. Haletky VMware Communities User Moderator, VMware vExpert 2009
Virtualization Practice Analyst
Now Available: 'VMware vSphere(TM) and Virtual Infrastructure Security'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|
Virtualization Security Round Table Podcast

Capacity Planner firewall

amvmware — Wed, 18 Nov 2009 15:00:46 GMT

Yes but based on my previous experience of trying to get CP working through a firewall i would do what vmware recommend and install another datacollector on the other side of the firewall - far easier than trying to convince the network security guy to enable access for the datacollector.

capacity planner - exception due to excessive paging

huwy — Wed, 18 Nov 2009 15:09:22 GMT

ok I changed the assessment to a CA and could find these options.
Thanks again for your help.

-H

Experiences of ESX and IBM Bladecenter

king@it.ibm.com — Wed, 18 Nov 2009 08:41:31 GMT

Have a look at this if interested: http://it20.info/blogs/main/archive/2008/11/14/162.aspx

As you can depict from one of the pictures you can "only" have 4 NICs and 2 SAS ports per each of the blade on the S. The way you configure those 4 NICs is totally dependant to what your goals/concerns/requirements/limitations are.

Ken Cline has a wonderful 8 parts series on NICs configuration for ESX hosts (http://kensvirtualreality.wordpress.com/?s=switch+debate) that is more than worth a read.

As Tyler suggested, if 4 is too low for you, the H is your best option. Consider that the S was really originally meant for remote sites / small shops, most of which would do well with even 2 NICs.

Massimo-

planning to upgrade to vsphere 4.0 and need help

kopper27 — Wed, 18 Nov 2009 00:59:19 GMT

is it really required to backup the existing datasbase for VC 2.x?

maybe
dump question but I think it is for just in case something goes
wrong...since I don't need the previous DB duting the installacion

iSCSI SAN questions & ESXi 4.0

weinstein5 — Tue, 17 Nov 2009 21:04:04 GMT

One other thing to keep in mind - to pay attention to the block size of the VMFS datastore because this will cap the size of the virtual disk you can store in the VMFS datastore - 1 MB - the max file size is 256 GB, 2 MB - 512 GB, 4 MB - 1024 GB and 8 MB - 2048 GB

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

ESX 4.0 New features/changes list compared to 3.5?

weinstein5 — Tue, 17 Nov 2009 19:41:38 GMT

This also might help - http://www.vmware.com/support/vsphere4/doc/vsp_40_new_feat.html

If you find this or any other answer useful please consider awarding points by marking the answer correct or helpful

Netapp 2020 SAS vs IBM DS3400

gstarr — Tue, 17 Nov 2009 15:44:48 GMT

Avg. Disk Bytes / Transfer - this is for IOPS (w2k3)?

last 7145,244

Average 16216,616

Minimum 4368,234

Maximum 113999,723

ok?

Storage, a Cloud storm or not?

sudiady — Tue, 17 Nov 2009 04:07:40 GMT

The general statement for Cloud Storage from an article "Cloud Storage For IT":

"Cloud storage is a broad term that typically defines a storage system as highly scalable, can be deployed for internal or external use and typically uses some form of clustered or grid based storage. It may or may not have a host of other features like geographic dispersion, retention capabilities and hardware independence.

When IT looks to cloud storage they are looking to solve the problems of meeting performance demands, dealing with massive data growth, overworked IT staffs and escalating storage costs. Each of these challenges will have a different weighting within the organization and that organization is going to select an architecture that best addresses them. The weighting of the importance of the above challenges will vary between organizations. If it is secondary data, then cost and scalability may be a bigger issue than performance. If it is primary data then performance may be the top priority. "

vSphere, Nehalem, and Citrix Xenapp

dnetz — Mon, 16 Nov 2009 09:58:39 GMT

Not much workload to mention, the cluster is running ~50 VM's on 4 hosts but with only a few test users yet. It was more a reflection on the step up in performance between Xeon architectures and how well ESX makes use of it.

VMWare ESX Licensing + Hosting

michael.custance — Sun, 15 Nov 2009 20:02:01 GMT

Did you get anywhere with this?

We are part of the VSPP but the licensing modle only works for enterprise VM's. We want to compete in this market place offering "value" machines. Problem is minium cost of a standard VM license is more than what our competitors are selling at.

I read that you could get an ammeded EULA apon request, so you are allowed to use licenses we have paid for, contacted VMware and they said this has now been removed "thanks!!"

Only workaround I can see in the EULA is tha tyou are allowed to provide administrative access if you are outsourcing. We would be provising administrative access to resource groups I guess this doesnt count though.

VMware are not really helping here, do they want people to use their products in this marketpalce or not? - I think they maybe getting a bit too greedy

.

Design Changes for vSphere?

Texiwill — Sat, 14 Nov 2009 14:42:00 GMT

Hello,

I would also look at the necessary steps and actions required to get to vSphere. Specifically versions of OS for vCenter, versions of SQL, etc. You may have more to upgrade than just your hosts. Ontop of this, you may want to reinstall over upgrade, so how can you temporarily use hostprofiles to make this easier, etc.

Just some thoughts on this.

Best regards,
Edward L. Haletky VMware Communities User Moderator, VMware vExpert 2009
Virtualization Practice Analyst
Now Available: 'VMware vSphere(TM) and Virtual Infrastructure Security'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|
Virtualization Security Round Table Podcast

Peak loads vs maximum observed

Rodge — Sat, 14 Nov 2009 13:57:46 GMT

Hi,

Within Capacity Planner I have some difficulties understanding the peak load value and maximum observed value. Can anyone help me.

"The peak load is determined by evaluating each metric over the most recent four weeks of collection and locating the hour of the day that has the highest average value. It is not the maximum observed value as any statistical analysis eliminates the high and low values from consideration.

Regards,
Rodge

extending vmdk over 1TB on NFS?

oreeh — Fri, 13 Nov 2009 23:47:41 GMT

IIRC the size limit for a VMDK (not a RDM) is / was 950GB.

Data Collector not getting Mem Inv data

kennyfranklin — Fri, 13 Nov 2009 18:47:46 GMT

Thanks again for the help guys.... changed inventory collection to registry only and upgraded collector and now I'm showing no more missing data anywhere. The help was much appreciated!

Kenny

Netapp FAS2020A vs Equallogic PS6000XV : Smoke and Mirrors

malaysiavm — Fri, 13 Nov 2009 17:55:51 GMT

To be honest, both storage are built from different architecture. Netapp is an unified storage support multi protocol and controller base storage architecture. Netapp come with Raid-DP which recommended by themself as to support 2 disk failure and improve performance over traditional raid 6.

Raid 50 is raid 5 + 0

Beside looking into it, you may need to consider for scaling, which Equallogic is shelf base scaling with additional controller per each shelf, which may end up require cabling management to be scale up when you add more disk shelf. Equallogic allow you to manage multiple disk shelf into single node and load balance across the disk shelf automatically.

Netapp architecture is more likely the controller base architecture, which always has a maximum supported number of disk for each different controller.

In term of performance, it always depend on the controller model,cache memory, spindles, throughput and others factor.

Replication wise, snapmirror from netapp and Volume replication from Equallogic does perform as it needed.

Dedupe is definitely help to reduce the primary storage consumption, but it will also depend what type of file will gain benefits from dedupe. typically virtualization, documents and images does benefits a lot from the deduplication. Maybe you should consider raid DP rather than raid 4 to obtain the best performance out from netapp boxes. You can always check with the netapp consultant on the best practices on thier storage configuration.

Craig

vExpert 2009
Malaysia VMware Communities - http://www.malaysiavm.com

VMview Desktop Migration

tom howarth — Fri, 13 Nov 2009 13:59:12 GMT

The replicas can be safely removed, once they are removed from the ESX servers View will cease to know about them

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points

Tom Howarth VCP / vExpert
VMware Communities User Moderator
Blog: www.planetvm.net
Contributing author on "VMware vSphere and Virtual Infrastructure Security: Securing ESX and the Virtual Environment”. Currently available on roughcuts

BladeCenter S and VI3

Macentosh — Fri, 13 Nov 2009 10:37:02 GMT

Could you provide me with the URL from IBM web site, please!

Thanks!

Req. documentation

Texiwill — Thu, 12 Nov 2009 17:58:55 GMT

Hello,

These documents are not publically available and most likely covered under NDA. You can however create your own without seeing these if you know enough about vSphere, etc.

Best regards,
Edward L. Haletky VMware Communities User Moderator, VMware vExpert 2009
Virtualization Practice Analyst
Now Available: 'VMware vSphere(TM) and Virtual Infrastructure Security'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|
Virtualization Security Round Table Podcast

Presenting 4TB LUN

Texiwill — Thu, 12 Nov 2009 17:19:11 GMT

Hello,

In order to use NPIV you must also create an RDM. NPIV is as Massimo stated just a way for SAN guys to get greater visibility at the moment. Eventually NPIV will be as you desire... I hope!

Best regards,
Edward L. Haletky VMware Communities User Moderator, VMware vExpert 2009
Virtualization Practice Analyst
Now Available: 'VMware vSphere(TM) and Virtual Infrastructure Security'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|
Virtualization Security Round Table Podcast

Good idea to use local Storage?

Texiwill — Thu, 12 Nov 2009 17:15:34 GMT

Hello,

Local storage is not a huge issue these days as you could get access to that local storage using software like Xtravirt Virtaul SAN or HP VSA. Or you could setup multiple virtual machines that replicate data between them.

The key is to determine how to establish your redundancy and as such ensure you have the necessary software or hardware to meet this plan.

Local storage is almost always your worst case business continuity case, so its not necessarily bad.

Best regards,
Edward L. Haletky VMware Communities User Moderator, VMware vExpert 2009
Virtualization Practice Analyst
Now Available: 'VMware vSphere(TM) and Virtual Infrastructure Security'
Also available 'VMWare ESX Server in the Enterprise'
SearchVMware Pro|Blue Gears|Top Virtualization Security Links|
Virtualization Security Round Table Podcast

Capacity Planner - Default reports

wathap — Thu, 12 Nov 2009 16:44:25 GMT

Hi amvmare/singy,

Some of the settings for CE reports could be change. The page per sec value could be changed.
You can change the threshold by modify the Scenarios that are used for the Optimization Reports. There are two scenarios and they are marked with ***

To edit the Scenarios, go to Analyze => Optimization Scenaios and click on the scenario names that have *** in front of the name.

Follow the wizard until you reach the page "Adjust System Thresholds"

Please note that there are two scenarios with *** and you should edit both of them.

http://communities.vmware.com/message/1400321

SAN on different IP Range

cstrachan — Thu, 12 Nov 2009 11:47:08 GMT

Answered it myself
thanks anyway

Best way to come up with a TEST plan to insure everything is working?

tjw82 — Thu, 12 Nov 2009 05:39:33 GMT

I was wondering if you ever found or came up with a test plan?

HA, V-Motion, DR options

Collin09 — Thu, 12 Nov 2009 04:55:19 GMT

The only concern if using a single ESXi is if there's a hardware failure. I'd rather have some sort of disaster recovery such as HA to link to another host server.

performance wise setup for SQL?

mlubinski — Wed, 11 Nov 2009 16:15:36 GMT

Thanks, but in posted links I didn't find any details as how VM should be configured for best performance. What I thought to see in such document would be something like:

if you set memory to 8GB, then set limit to 8GB, and reservations to 50% of limited memory

inside Windows 2008 set pagefile to automatic (or whatever is recommended)

on VM configuration set to place swapfile on host swapfile datastore

etc, etc

Maybe someone could share about their internal configs, that perform very good on VMware with SQL (and best if this would apply also to storage over NFS)

If you found this or any other answer useful please consider the use of the Helpful or correct buttons to award points

AMD IOMMU versus Intel VT-D

Photubias — Wed, 11 Nov 2009 12:56:54 GMT

We first need to think why VMotion is not possible with Intel VT-d:

VT-d (currently) is the option to assign a PCI device (or in the future with SR-IOV to assign a PCI Function) to a single Virtual Machine.
This means that a VM has full control over that full device and none of the other VMs can access that device. (Just as in a native situation).
However, this is actually a stap 'away' from virtualization (separating hardware from software): when a VM is VMotioned, this would be the same as just pull out a PCI device out of a native computer, resulting in a BSOD at best or even hardware damage. This is the reason why VMotion is not a good idea with Intel VT-d.
AMD IOMMU is the same technique so I image VMotion won't work either.
AMD IOMMU support in ESX is only experimental, however.

One last comment:
I understand it to be like this (*please* correct me if I'm wrong):
Intel VT-d and AMD IOMMU are Intel and AMDs solutions to Single Root I/O Virtualization (SR-IOV) as suggested by the PCI-SIG group.
And SR-IOV is actually this: assigning a PCI function to one single VM. Notice I say: PCI function and not PCI device; some PCI Devices can have multiple VFs (Virtual Functions) and one VM can be assigned to 1 VF, this would not be a problem for VMotion however.
This must by supported by the chipset (Intel VT-d on a Intel Chipset), by the Hypervisor (ESX has that support) AND by the PCI device, as far as I know, LSI is the first to have an SAS Controller with multiple VFs. Intel has NICs with these functions, namely Intel VMDq. Also broadcom has allready (10Gb) adapters wich are SR-IOV capable.

VMFS vs RDM (vs VMDirectPath and other solutions)

AndreTheGiant — Fri, 18 Sep 2009 15:15:15 GMT

VM disk type

There are different solution for implement a VM disk:

(native) vmdk over VMFS datastore (see also the different VMDK virtual disk type)
(native) vmdk over NFS datastore (vmdk format usual is thin)
virtual RDM (not for NFS datastore)
physical RDM (not for NFS datastore)
NPIV RDM (only for FC storage)
(native) guest iSCSI
direct with VMDirectPath I/0 (see also VMware VMDirectPath I/O)

How to choose virtual disk for a VM, iSCSI initiator in ESX, iSCSI initiator in OS or RDM

Performance difference

http://www.vmware.com/files/pdf/vmfs_rdm_perf.pdf
http://www.vmware.com/files/pdf/performance_char_vmfs_rdm.pdf
http://www.virtualization.info/2008/02/benchmarks-vmware-vmfs-vs-raw-disk.html
http://it20.info/blogs/main/archive/2007/06/17/25.aspx

Virtual Compatibility Mode Compared to Physical Compatibility Mode
http://pubs.vmware.com/vi35/server_config/wwhelp/wwhimpl/common/html/wwhelp.htm?context=server_config&file=sc_adv_storage.12.6.html

VMDK disk vs RDM disk

VMFS vs. Physical Disk performance
vmfs or raw for vms? which to choose
Question re RDM vs VMDK for high performance
RDM vs. VMFS...again...
vmdk vs RDM for large disk
VMFS vs raw data mappings in windows 2008.
Difference between a VMDK disk and Virtual RDM disk

Usually the "keep it simple" approach is the best choice... and vmdk over VMFS are very simple...

But anything that adds an "additional layer of abstraction" must therefore create a (no matter how small) overhead. For this reason when you have mid-range to high-range I/O on a disk, the RDM will perform a little better than a vmdk inside a VMFS.

For any database over a reasonable size I could be a good idea to create the database drive as an RDM - first to remove any I/O hit, and second to give the flexibility to access that RDM-ed LUN on a physical box without the need to 'convert' or V2P it.

Remeber one thing that a lot of people forget... the VMFS mounts from your SAN come to the hosts over the SAME fibre infra as the RDMs (same switch, same HBA etc) - if one VMFS-hosted VM is leech-whoring all the FC bandwidth then having your database be on an RDM is not going to help matters, is it?

VMDK vs VMDirectPath I/O

http://kb.vmware.com/kb/1010789 - Configuring VMDirectPath I/O pass-through devices on an ESX host
VMware VMDirectPath I/O

Actually VMDirectPath I/O it the best solution for VM with very high I/O, cause the performance are like in "native" mode.
But you loose a lot of the advantages of the virtualization: no VMotion, no backup, no cold migration between ESX, ...

VMDK vs native iSCSI

Re: iSCSI Performance
Creating VMFS on multiple internal SCSI disks

This solution could be very simple and "natural" in a iSCSI environment.
But remember that VM with a guest iSCSI cannot be protected with a backup solution for virtual environment, cause VCB, VDR or similar program can not "see" the iSCSI disks...

Any known issues when P2V'ing specific Enterprise servers?

BenConrad — Wed, 11 Nov 2009 04:39:39 GMT

A few thoughts:

Watch out for weird boot.ini entries, I've had a few nostandard entries hang the VM post-P2V (can't remember what entries, it was 2 years ago).

Don't underestimate how much work a relatively new server (2 x 4) core server can do. An 8-core server at 5% CPU utilization will certainly result in a busy uni-proc VM.

Don't allow the VMs to use BusLogic adapters (very low max queue depth)

Make sure to change the HAL to uni-proc for 1vCPU VMs.

Using Converter 4.0.1 I've seen a lot of BSOD's on 'reconfiguration' lately, I think this can be fixed by upping the system PTEs or performing the reconfig separately.

Ben

Network Outage?

rbmadison — Tue, 10 Nov 2009 15:03:00 GMT

Thanks for confirming! I appreciate everyones comments!
Rick

EVA 4400 VRAID 1 or VRAID 5?

digitaex — Tue, 10 Nov 2009 13:43:07 GMT

We do vraid 5 for all OS and APP volumes. Depending on the type of i/o we do some vraid1 on sql data-logs-tempdb volumes.

750GB vr5 OS Volume (10x50G OS only with room for snapshots)
750 GB vr5 APPS Volume (10x50GB for applications)

iSCSI Switch Recommendations

hutchingsp — Mon, 09 Nov 2009 16:38:12 GMT

Brilliant, thanks for posting that - very useful and hopefully the "missing link" should we settle on the Equallogic.

VM Density

king@it.ibm.com — Fri, 06 Nov 2009 20:45:53 GMT

I think that ultimately the scenarios where the unlimited thick box should be "unthicketed" are very niche compared to scenarios where you want it to be "thicketed" and perhaps reserve 50% or 33% (examples) of the assigned + shares to do the VMs ranking.

At the end of the day you want the ultimate "limit" to be the pRAM installed in the box vs a hard set limit on specific VMs (unless you are 100% sure that you won't use more than the reserve you specified (but if you are 100% sure why thicking it anyway as it's the end-user software that is going to limit itself).

I totally agree that this needs to be taken with a grain of salt. There are applications (SQL for example) that will tend to use all the memory that they see no matter what (and that's the assigned.... not the reserved). If all VMs are using software that behaves like this you'd better not even overcommitting memory...

Massimo.

why is storage not handled as a ressource?

amvmware — Fri, 06 Nov 2009 15:40:56 GMT

I am not aware of a feature in vSphere that will set quota limits full stop on a datastore - so it must be.

Static Workstaiton VM's vs. Automatic Allocation Pool's

Anton V Zhbankov — Fri, 06 Nov 2009 07:48:59 GMT

In case of View:
1) All users connect to View server, they don't have to remember different IPs/names of their VMs.
2) View creates VM automatically for each new users according to rules
3) You can setup additional View server in DMZ allowing users to connect from outside
4) Connection is secure, rdp over ssl
5) View allows you to map local USB devices to VM

If you have 5-10 users then yes, VIew is not really helpful. But 100, 200, 500 VMs are a great headache.

---
MCSA, MCTS, VCP, VMware vExpert '2009
http://blog.vadmin.ru

contents of Capacity Planner system inventory report

Limeres — Thu, 05 Nov 2009 20:14:39 GMT

Hi,

Thank you for comments.

I have not ran the Test inventory during install...

I will start the full inventory export tomorrow.

Thanks to All!!

I will post the result of my proof.

Regards!!!!

Patricio

HP Blades and MSA 60 ..

amvmware — Thu, 05 Nov 2009 16:25:20 GMT

I would agree with the previous comments.

Blades - unless you have a requirement to fully populate a blade chassis then it is not a cost-effective option as opposed to purchasing a couple of DL360 or 380 G6.

Storage - If they are small then look at the MSA 20xx iSCSI or FC solutions - I think a NetAPP would be to expensive - once you total up all the additional options you want a NetApp is not a cheap storage option for an small customer.

Server 2008 Clustering in ESX 3.5 Update 3

sunilhaste — Thu, 05 Nov 2009 12:07:18 GMT

Hi Amit,
Will you be able to give me guidance as to how you did this. I am a novice in this area and I have to do exactly what you have done. any help in this regard or any pointer will be very helpful indeed.

invalid vlan tag error

TimOudin — Wed, 04 Nov 2009 21:47:05 GMT

Same harmless logging here, IBM HS21 with Cisco switches.

Cheers

Capacity Planner system thresholds

wathap — Wed, 04 Nov 2009 18:24:23 GMT

Hi Rob,

I am not sure myself. I did both deactivate and isolate them.

Help on picking a new San

jjkrueger — Wed, 04 Nov 2009 16:36:33 GMT

In choosing a SAN, you need to think about what the SAN is going to need to do. Take a look at your servers, figure out how much storage throughput is needed and how many I/O Operations per Second (IOps) they need. What I generally find is that IOps are the bottleneck, rather than throughput.

You have to find that out about every server you're planning to virtualize, and you can then begin planning your storage around that. I don't know the specific numbers on the NetApp or Hitachi box, but you should be able to go back to your storage vendors and say "I need to be able to support X VMs with Y IOps and and Z throughput" and they should be able to tell you which boxes can provide that performance. The performance requirements will ultimately determine what kind of disks are used in the array, and how the back-end LUNs are carved up. You need to think about VM placement on the LUNs, and ideally performance profiles for each of the VMs so you can place high-IO VMs that have different peak times on the same LUN for overall performance.

Since you're looking at database servers, I would also talk to your storage vendors about caching mechanisms in the storage arrays and how that will impact design choices - different arrays do caching differently.

Since you're concerned about the future, it would be a good idea to include extrapolated performance needs for the lifecycle of the device so that you don't get into a conundrum with buying a box that fits your needs today, but not tomorrow. Typically, a 3 year extrapolation is good, as most storage vendors seem to like to shuffle kit on a 3-year plan.

I know that doesn't exactly answer your question, but I hope it gives you some direction to help with your decision.

-jk

iSCSI SAN, ESX4 Datastore Sizing Best Practices

daveclaussen — Tue, 03 Nov 2009 19:53:09 GMT

Thanks for your help. Now I have a plan!

Looking into migrating Exchange and SQL from Physical servers to VMs...thoughts?

amvmware — Tue, 03 Nov 2009 19:40:50 GMT

Tommy

You might find the following useful.

This is my rule of thumb.

The virtual machines will be mixed on the datastores to maximise the storage efficiency and I/O efficiency - mix low I/O virtual machines with high I/O virtual machines on the same datastore.
A maximum of 12 - 16 virtual machines per VMFS datastore.
A maximum of 16 VMDK files per VMFS datastore.
One VMFS volume per storage LUN.
At least 15% to 20% of a VMFS datastore should be left as free space to accommodate requirements such as virtual machine swap files and snapshots.

Netapp ESX host utilities - any issues?

maxel — Mon, 02 Nov 2009 23:54:50 GMT

Is mbralign only useable with NetApp Storage Arrays or with other FC SAN Storage Devices too? Is there a download of this tool without being a NetApp Customer?

VSphere Capacity Planning

vmproteau — Mon, 02 Nov 2009 21:14:28 GMT

Ignoring CapacityIQ which I understand isn't yet supported in VSphere currently, I'm trying to develop a logical calculation to estimate capacity in a particular environment. We use an N+1 sizing strategy so, for 2-Quad Core, 32GB memory Hosts, we would max our N+1 memory ceiling and buy another Host before CPU would ever be a concern.

We are now looking at the same amount of cores (2-Quad Core) but increasing to 96GB of memory (HP DL380 G6) so now I'm more concerned with CPU performance. I have heard of some generic VM/Core best practices (4-6 for ESX 3.5) (8-10 for ESX 4.0) however, I'm looking for a more accurate calculation.

Can anyone tell me what metrics I would need to look at and perhaps what calculation I would make to best estimate what kind of CPU capacity I safely have in a paticular Host/Cluster. I'm not even sure if I need to be looking at average MHz over time, vCPU/Core, VM/Core, etc.?

I know there are tools for getting this information like CapacityIQ but, I am trying to "right size" our future Hosts hardware standard.

Evaluating HP, Dell, IBM Solutions for small VMware deployment

king@it.ibm.com — Mon, 02 Nov 2009 09:07:58 GMT

Chiper1,

the only thing I can say is that the BC-S solution has been built with that specific architecture in mind. It is by far the "most elegant" of all.

http://it20.info/blogs/main/archive/2008/11/14/162.aspx

Having this said your scenario is pretty entry-level and minimalist (so to speak) and you won't be wrong if you decide to go with the BC-S alternatives.

Massimo.

vMotion Porgroup & Serice Console PortGroup can have the same IP ?

shatztal — Mon, 02 Nov 2009 05:47:27 GMT

these reasons i know i thought somthing more efficent maybe . thanks again for the help

Windows Storage Server 2008 in ESXi

wibni — Mon, 02 Nov 2009 03:15:39 GMT

Hi,

Thanks for your replies!

Odysey wrote:
Simply partition the local disks as VMFS and then allocate VMDK's to the guests, much better option then using network storage for local VMs.

I think that's what I've done now. I've got 2 datastores. 1 with RAID 1 containing ESXi and the other RAID 5 containng my VMs.
I'm not completely sure what you mean by "allocating VMDK's to the guests"? Do you mean to create a blank virtual disk and allocate the new disk to a VM somehow - just like adding an additional hard drive to a physical server?
How would I do that? Just create a VM with no OS on it in the vSphere Client?

Equallogic Feedback (and vs. Lefthand)

Peter.Grant — Sun, 01 Nov 2009 18:22:34 GMT

I'm stating this from memory so could be wrong but I think u can span a volume. With LH u creat a cluster with one or more nodes in it. When u create a volume it's a volume within the cluster so it should be spread amount the nodes. Could be wrong but pretty sure that's how it works. Choosing no replication , 2-way etc only specifys the network RAID level hence redundancy

-------------------------------------------------------------------------------------------------------------------
Peter Grant
Virtualization Consultant
Xtravirt.com

Please award points if you found this helpful

ESXi 4.0 on SuperMicro MaxServer 6015B-TB

RDellimmagine — Sun, 01 Nov 2009 05:42:41 GMT

Thread moved.

Robert Dell'Immagine, Director of VMware Communities

ESX IP Addresses Migration

Troy Clavell — Sat, 31 Oct 2009 19:59:19 GMT

http://communities.vmware.com/thread/239965

VMware Data Recovery or VCB?

AndreTheGiant — Sat, 31 Oct 2009 05:32:03 GMT

Can i say that the better option is that using 3rd party application which utilize vStorage API ?

Right...
But also the backup program must be a good backup program

Andre

What is the benefit in enabling CDP (Cisco Discovery Protocol)?

contra422 — Fri, 30 Oct 2009 18:53:41 GMT

Its just another way to gather useful information - the port details are useful if you need to trace the cable and the network team boys are happy because they can confirm the configuration they are applying is on the correct server/port.

What hardware do you most wish ESX supported?

mike.laspina — Fri, 30 Oct 2009 16:02:29 GMT

IBM Branded Qlogic CNAs 42C1800 (qla8142)

http://blog.laspina.ca/
vExpert 2009

Win2003 Multiprocessor to Uniprocessor HAL

NYSDHCR — Fri, 30 Oct 2009 15:46:56 GMT

I'll try applying SP2 again, and I'll let you know...

Virtual SMP Processing

AlbertWT — Fri, 30 Oct 2009 06:41:28 GMT

OK, great thanks for the suggestion Peter.

Cheers.

Kind Regards,
AWT

iSCSI SAN Solution

Peter.Grant — Thu, 29 Oct 2009 22:04:34 GMT

I've never used it but it seems popular by some. All I've used for iSCSI is LeftHand solutions, they worked fine for me and did the job. I've used them on a few designs including enterprise deployments.

-------------------------------------------------------------------------------------------------------------------
Peter Grant
Virtualization Consultant
Xtravirt.com

Please award points if you found this helpful

Is DRS possible with no shared datastorage ?

Peter.Grant — Thu, 29 Oct 2009 16:40:35 GMT

If you want to have DRS and VMotion and you don't own a SAN you can use a software SAN for free. Go to the VMware Virtual Appliance Marketplace and type in SAN. You can then setup any storage you want, configure your ESX hosts to connect via ISCSI and then you have your shared storage.

Other products like LeftHand networks also have a virtual SAN you can trial but only for 90 days. The Xtravirt SAN which we sold to PHD Virtual is free.

-------------------------------------------------------------------------------------------------------------------
Peter Grant
Virtualization Consultant
Xtravirt.com

Please award points if you found this helpful

Backup Virtual Machine to External Hard Drive

petedr — Thu, 29 Oct 2009 03:38:54 GMT

Thanks Peter for mentioning esXpress

Pete@esXpress

www.phdvirtual.com, makers of esXpress

Small vShere 4 san configuration

Peter.Grant — Wed, 28 Oct 2009 22:49:12 GMT

I'd say you can more than likely get away with 4 NICs per host for what you've described (assuming 1 Gbs NICs). If it were me I'd use 4 in the following config:
2 x vSwitches
vSwitch0 for iSCSI with 2 NICS

Configure 2 x VMkernel ports for this switch and bind each to a specific NIC. Do this by setting each NIC to Active on a respective NIC (in the NIC teaming section of the vSwitch0) and Unused for the other NIC
At the SSH command line use the command to bind the iSCSI s/w initiator to the NICs using the following commands

esxcli –-server <servername> swiscsi nic add -n <VMkernelportname> -d <vmhbaname>

Don't worry it's all explained in much better detail here:(iSCSI config guide pg 30) http://www.vmware.com/pdf/vsphere4/r40/vsp_40_iscsi_san_cfg.pdf

Then configure vSwitch1 for Service Console and VM traffic in an Active/Active Configuration. (1 port group for each on the same vSwitch)

There's no perfect answer but I dont think you'll have any problems with 4 x 1GB NICs when you only have 8 VMs. Keep it simple is always good.

Your C: and D: config again seems fine, and with this small number of VMs just create one datastore and keep all files in the same folder as the VM. This is assuming you dont have any high IO VMs.

I'm not 100% sure about the dual controller question but if you can just have one RAID set with dual controllers then do that, keep it simple.

Also check out this post:
http://virtualgeek.typepad.com/virtual_geek/2009/09/a-multivendor-post-on-using-iscsi-with-vmware-vsphere.html

Pete

-----------------------------
If you find this helpful then please award points...you know you want to!

Server 2008 x64 on ESX 3.5

dnetz — Wed, 28 Oct 2009 08:04:10 GMT

Same here, we've been running a bunch of domain controllers on Win2k8 x64 on ESX 3.5 U4 with no apparent problems.

vCenter in VM vs HA

Anothai — Wed, 28 Oct 2009 07:19:04 GMT

I have configured vSphere4 same as your hardware infrastructure. My vCenter is running in VM. I have same experience that HA feature is not function when i poweroff one of ESX server or pullout network cables from one of ESX server. Unfortunately, i fix my issue by installed vCenter servicepack 1 due to vCenter bug for few IP address range. Please see more details in

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1013013

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1013013

Regards,

Anyone seen this news: VMWARE saves the vSphere Enterprise Edition?

wila — Tue, 27 Oct 2009 20:10:08 GMT

Well it all depends on what you read
Here's another article about it: Stay of Execution for vSphere Enterprise
and I agree with you that it was a pretty important announcement, but it was announced almost a month ago already...

--
Wil
_____________________________________________________
VI-Toolkit & scripts wiki at http://www.vi-toolkit.com

One ESX 4 Server connected to EQ 3TB over iSCSI, How would you do it?

Chuck8773 — Tue, 27 Oct 2009 19:31:09 GMT

VMDK, RDM, Storage Direct are all very valid options.

VMDK is ideal for many smaller files. As the file size grows, you should consider switching to RDM. Storage Direct gives you benefits in application consistent copies. VMDK and RDM give you benefits in mobility. You shouldn't be moving these files around very much, but when you want to, it is very nice to have on this side.

For us, application backups are handled with a separate process. We are opting for VMDK files for all data. This sets us up well for things like changing storage groups, using a different SAN technology, and maintaining a hot site. If ESX knows about all of the storage needs of the VM, it is easier for ESX to handle failing to a hot site.

Evaluate all options. All three methods have their place.

Charles Killmer, VCP4

If you found this or other information useful, please consider awarding points for "Correct" or "Helpful".

P2V Linux init level

AndreTheGiant — Tue, 27 Oct 2009 18:23:00 GMT

You can use single mode, but then you have to enable network and also ssh.

Andre

What 10GB iSCSI SAN's are available to buy NOW

Box293 — Mon, 26 Oct 2009 22:28:20 GMT

Yes, the enterprise products are starting to become available.

VCP3 32846
VSP4 VML-306798

Deploying exchange 2007 in ESXi4 – storage spindle configuration?

amvmware — Mon, 26 Oct 2009 20:49:42 GMT

The only thing that i can see would impact Exchange performance is if the exchange server is being heavily utilised by all 36 users - if you only have a few heavy users then it should not be an issue to virtualise exchange and use VMDK's on a single or multiple datastores.

Other things i would do.

1. Setup performance counters so you know how it is currently performing and it provides a baseline to performance when physical. - Usual things, CPU, Memory, disk and exchange counters - see MS doc for relevant Exchange counters.

2. See the following VMware article on virtualising exchange - http://www.vmware.com/resources/techresources/10021

Storage VMotion Simplified

meistermn — Mon, 26 Oct 2009 20:46:56 GMT

If you need to move many vm's from one array to many, there to nice solutions:

1.) From get admin Evacuate a Datastore

http://get-admin.com/blog/?p=779

2.) Me with great help from Lucd

http://communities.vmware.com/thread/236684?tstart=45

DHCP IP-Adress for the ESX Service Console a good plan ?

meistermn — Mon, 26 Oct 2009 20:34:11 GMT

The vmworld presenation "*TA3195 ▲ VM Stateless ESXi: Scalable Rollout and Management of* Virtualized Hosts" cleared it for me

new cisco 3750 network design and vSphere4

habibalby — Sat, 24 Oct 2009 17:46:31 GMT

Hello,

Yes, i forgot to mention that, it's better to connect one Physical nic on pSwicth1 and the other on pSwitch2

Best Regards,

Hussain Al Sayed

If you find this information useful, please award points for "correct" or "helpful".

VMmark - understanding what's better - DL380G6 or DL385G6?

markzz — Sat, 24 Oct 2009 16:56:07 GMT

What rmagoon said is so very true.

AMD I'm sure will soon enough get back in the ball park but for now Intel are kicking their behinds.

CPU is very unlikely going to be the limiting resource..Memory, Storage IO, NIC IO, usually falter before CPU

HP DL380 G6: ESX Host Memory or CPU

vmproteau — Sat, 24 Oct 2009 00:05:50 GMT

We have a primarily HP shop where our ESX Hosts are HP DL380 G5, Quad Core, 32GB RAM. Generally memory is the deciding factor on when to consider adding Hosts. CPU isn't really considered because our VM/Core ratio is relatively small.With VSphere and lease replacements looming we are likely looking at HP DL380 G6 Quad Core, 64-72GB RAM. Those familiar with the G6 may know that memory population can greatly affect speed where 1333 MHz memory can be reduced to 800MHz speed depending on configuration.

To maximum memoy speed at the proposed capcities, we'll have to spend 10X on 8GB DIMMs. If we did that we'd probably look at less of a processor. I am concerned with the larger boxes because our VM/core number will certainly increase so my gut tells me to get the fastest processors I can afford but, I have no evidence to back that decision.

If I had to sacrafice one for the other should I go with faster CPU or Memory speeds. I do realize the VM population would factor in. We have a shared environment so server role runs the gamit but, they are generally Windows 2003/2008. Is there any guidance or evidence to help determine if the benefits of faster memory outweigh CPU speed concerns or vice versa?

On a related note, anyone know why the HP DL300 series G6 Intel offerings only Quad Core while the AMD offers Hex. Are Intel Hex in the works?

VMware P2V Converter OS license

TobiasKracht — Fri, 23 Oct 2009 15:56:19 GMT

Mark my post as useful or correct, please, if so

StarWind Software R&D

Moving guest OSes from a failed host to another host w/o vMotion..

shandbrus — Fri, 23 Oct 2009 15:34:04 GMT

perfect...both very helpful answers...thanks a bunch. i new it'd be easy like that. (yes the hosts can see all the datastores)

thanks again to the both of you

shan

Store or not to virtual disk with VM?

Wimo — Fri, 23 Oct 2009 13:40:53 GMT

We've never had any problems putting vdisks in different datastores. We (almost always) create our disk from the SC with vmkfstools, since that allows you to name them (a feature still lacking from VC). It's a lot easier to tell what's what with servername-e.vmdk, servername-g.vmdk etc than servername_1.vmdk, servername_2.vmdk.

On the other hand, we do not use snapshots for server VMs, so I can't speak to any other complications from that.

Best approach for a Windows WebFarm

alnapp — Fri, 23 Oct 2009 13:35:22 GMT

Thanks for the info hstagner

I thought that was the case with VDI, but wanted confirmation.

I take your point aboutmemory consolidationetc and am already doing that, but is there any disk resource consolidation that could be done (shared OSdisks, etc.)?

Cheers

Periodic clean up of inactive VMs

crowhurst — Fri, 23 Oct 2009 12:58:49 GMT

Can someone point me in the direction of a canned report to identify VMs that are suspiciously inactive? I am thinking something along the lines of very low levels of cpu and disk activity for a prolonged period. I am trying to periodically spot those VMs that are basically just an idle OS and to ping their owners to see if they are still needed. Thanks

Vsphere - MD3000i configuration

Rob.Bohmann — Fri, 23 Oct 2009 02:48:57 GMT

Hi ChrisVan,

I just started using one of these units this week, though it has the 300 GB SAS drives. With 7 disks in a RAID 5 set for the first disk group I get about 1.6+ TB usable space. They are super easy to configure, even I could figure it out ;>) without difficulty - and I am not a "storage guy".

If the servers you want to run on this are already running somewhere else, I would advise you to measure them for a few days or a week to see what the load is - or ask your dev team to stress test them for an afternoon to get some idea or what kind of load they may generate. My first test was a clone operation today, I was streadily pushing about 120-130 MB/sec combined read/write. This was an empty lun with only a couple VM's powered on within the disk group which was doing very little otherwise.

I think if you put the each SQL on a separate lun and spread them across a couple of disk groups you should be fine - thats just a guess not knowing what kind of load they generate, and what else will be co-existing with them, but from the sound of this it does not seem like a bunch of heavy hitters. I hope to be able to perform some more thorough tests in the future, but it probably be after you have to make a decision.

What are you using for your current server with local storage? How many spindles and what kind of controller/disks? Unless you have some monster setup I cannot see how this would be a lesser perfromer.