Performance Problems W2003, red indicators on Task manager

[Ralf-B]

After we migrate our File Server to a Virtual Server (we installed a virtual Server and connected the Data Partitions as RDM (phys)

Backup is very slow, Some USers says access to the file is to slow too.

We have got some red indicators on Task manager and there are NO programms which are consuming so much processor time,

What does this red indicators means, have can i found out a solution for this Problem?

Thanks for any help.

[Rumple]

Lets start from the top...

#1 - Virtual Server - are you talking VMWare server, Microsofot Hyper-V, or an actual vSphere Environment?

#2 - Exactly what kind of network gear are you running (cisco, hp, dlink, dell,etc)

#3 - What is running your storage. Obviously some type of SAN if you are using RDM's

#4 - How are you connecting to the storage (ISCSI, FC?)

[Ralf-B]

sorry,

esx 3.5

network, backbone is cisco

ibm ds 4700

FC

[Ralf-B]

sorry,

esx 3.5

network, backbone is cisco

ibm ds 4700

FC

I'm trying to upload a screenshot from the taskmanager of VM, but is not possible.

You can look the Screen shot

http://i43.tinypic.com/2d0z3gj.png

[Rumple]

If the task manager in the vm is running at 100% then it means it is not getting access to enough cpu time on your esx host.

How many cores are in your esx host and how many multi cpu vm are you running.

You want to make sure that every vm is configured as a single cpu and only increase them to 2 cpu if necessary.

If you did a p2v I would suspect that server is running at least 2 cpu's...you's need to change the hal to a uni type and then shutdown and drop cpu's down to 1.

[Ralf-B]

my customer configured this vm (no p2v) with to vCPU. When i see it i changed this and the other VMs to 1 vCPU but the HAL is at the moment for multi CPUs (W2003) and it is not easy or not supported to change this one CPU HAL or ? Because i read something about this on Microsoft website.

The host has 2 sockets with 2 cores. On this host, are only 3-4 VM not more.

What does this Red indicators mean ?

Thanks

[Rumple]

The red indicator is that ESX is unable to give that VM some CPU time because it is busy giving other VM's access to the CPU's. If you have 4 cores total and 3-04 VM's with single CPU you will be fine..but if the Operating system thinks there is another CPU, it will send commands to the nonexistent CPU and will wait for the command to finish (which it cannot because there is no physical CPU to process the command).

You can easily change the HAL of a server down to Single CPU.

Step 1 - Put the VM into snapshot mode

Step 2 -

Go into Device Manager, Computer, and right-click on the ACPI Multi-processor driver and choose Update Driver.

Pick "Install from a list" and then click on Next

Select "Don't search. I will choose the driver to install" and click on Next

Choose the ACPI Uniprocessor Driver.

Reboot.

If everything comes up fine, make sure to remove the snapshot.

Make sure ALL other VM's also have a matching number of CPU's and HAL's configured...

[Ralf-B]

i know this way how to change the HAL but i thought this way is not supported or am i thinking wrong.

I see the 2 CPU after my customer installed and configured the VM.

Ok i will try this,

thanks for your help

[a_p_]

Even though I agree with Rumple, the first thing I would do in this case was to shut down some of the other VM's to see whether or not this makes your file server run with "normal" CPU usage. (If this is possible in your production environment.)

André

[Ralf-B]

i moved all the other VMs to the other Host. This VM was the only one.

During an offday, there wa no users on the system, only the backup job was running, and the same Problem with Taskmanager.

My Backup Job runs with factor 1000 (backup Program indicator), normally this runs with 10.000 and more.

I dondt know what this VM has a Problem.

[a_p_]

Please allow me some questions:

- Are your ESX servers and the components (HBA's, NIC's) in a supported state (VMware HCL, BIOS/Firmware versions, ...)

- Did you follow/review the IBM RedBook VMware Implementation with IBM System Storage DS4000/DS5000

- How did you setup the virtual disks for the file server VM (LSI/Buslogic, physical or virtual compatibility mode)

- Are the OS and RDM disks on separate SCSI buses. e.g. OS on SCSI(0:0), RDM on SCSI(1:0)

André

[Ralf-B]

yes all supported Versoin.

The OS Disk is virtual disk

The Data disk is RDM Physikal Mode

We used the default controller for W2003 OS

[a_p_]

What about the SCSI ports, the HDD's are connected to?

Please provide the vmx file, if possible.

André

[RParker]

i know this way how to change the HAL but i thought this way is not supported or am i thinking wrong.

Windows XP and Windows 2000 seemed to have issues reverting from Multi to Uni, and some point (much like the comment that Bill Gates ;supposedly made, but no proof he ACTUALLY said; that 640k should be enough for anyone) things get miscommunicated or people ASSUME that's what happened.

In this case there are KB articles on MS site that freely tells you how to change it, so the "difficulty" here is just doing it correctly. It's not hard, its very easy, it just takes time. MS has made it EASIER to change it, with Windows 2003. The caveat is SP1 vs SP2. SP1 you had to use the OLD method (XP/2000) you put the OS in repair mode, and reinstall the OS to let it change the HAL. SP2 is simply going into device manager, change the CPU driver to uni, and power OFF the machine (reboot won't do it, you have to power it off first).

So depending on who you talk to, this process is either impossible (because people are inherit lazy, and just don't want to do it, so THAT's where most of the confusion comes in, they basically tell people it can't be done, because they don't want spend the effort to explain, or people can't do a KB search on their own using google or the MS support site) or details about HOW to do it precisely are not conveyed properly.

So changing the HAL IS supported, HOW you do it is up to you, you can either read the KB articles or "guess" how to do it, if not done right you will get performance issues much like you have now. Running a multi CPU on a single CPU is LESS supported than changing the HAL. So actually you are running an unsupported configuration NOW.

Change the HAL, and the problems go away.