VMware FT & your HA slot size

Something you might need to consider when using the VMware Fault Tolerance (FT) feature is how it can have a very big impact on the quantity of machines your cluster can run.

When you enable FT for a VM, a reservation will be made to the size of the memory associated with that VM. So if your virtual machine has 4GB vRAM then the reservation will also be 4GB.

This is all fine and dandy provided it’s not sitting within a VMware cluster with HA enabled using the admission control policy ‘number of host failures to tolerate’. Because this cluster setting uses a slot based algorithm to decide how many virtul machines can be powered up based on the largest memory and CPU reservation, having a huge memory reservation set for an FT protected VM will reduce the number of slots available.

Best thing to do is either change the slot size which you can achieve through the advanced settings or change the HA admission control setting to use ‘Percentage of cluster resources reserved as failover spare capaciy’.

A VMware kb article: http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1012602 provides more info on changing the slot setting.

 

DRS/HA maximums per host change with vSphere 4.1

You may or may not be aware that VMware have always stated a supported maximum number of virtual machines per host in HA & DRS clusters. Because the number changed significantly between an 8 & 9 node cluster this was often used as a contributing factor for sizing.

However, as of vSphere 4.1 the limits have been increased and are no longer linked to the cluster size.

Below are the details for v4.0, 4.0 U1 & 4.1

VM’s per host in an HA cluster with 8 or < hosts  (4.0 = 100) (4.0 U1 = 160) (4.1 = 320)

VM’s per host in an HA cluster with 9 or > hosts (4.0 = 40) (4.0 U1 = 40) (4.1 = 320)

VM’s per host in a DRS cluster  (4.0 & 4.0 U1 = 256)

Maximum VM’s in an HA/DRS cluster (Powered on) (4.0 & 4.0U1 = 1280) (4.1 = 3000)

vSphere Update 2 released

I’m sure this news spreads quicker via the blogs than VMware’s official announcements but in-case you’ve missed it, there’s a new version of vCenter & ESX available today.

There’s a bunch of new O/S customisation enhancements for (XP SP2 64-bit, Win2k3 SP2, SLES 11, SLES10, RHEL 4.8, 5.4 & 5.5, Debian 5.0 R1 & R2)

The usual array of bug fixes, plus a few interesting enhancements to ESX including a nice addition to resxtop (see below)

• Additional Guest Operating System Support— ESX/ESXi 4.0 Update 2 adds support for Ubuntu 10.04. For a complete list of supported guest operating systems with this release, see the VMware Compatibility Guide.
• Enablement of Fault Tolerance Functionality for Intel Xeon 56xx Series processors— vSphere 4.0 Update 1 supports the Intel Xeon 56xx Series processors without Fault Tolerance. vSphere 4.0 Update 2 enables Fault Tolerance functionality for the Intel Xeon 56xx Series processors.
• Enablement of Fault Tolerance Functionality for Intel i3/i5 Clarkdale Series and Intel Xeon 34xx Clarkdale Series processors— vSphere 4.0 Update 1 supports the Intel i3/i5 Clarkdale Series and Intel Xeon 34xx Clarkdale Series processors without Fault Tolerance. vSphere 4.0 Update 2 enables Fault Tolerance functionality for the Intel i3/i5 Clarkdale Series and Intel Xeon 34xx Clarkdale Series processors.
• Enablement of IOMMU Functionality for AMD Opteron 61xx and 41xx Series processors— vSphere 4.0 Update 1 supports the AMD Opteron 61xx and 41xx Series processors without input/output memory management unit (IOMMU). vSphere 4.0 Update 2 enables IOMMU functionality for the AMD Opteron 61xx and 41xx Series processors.
• Enhancement of the esxtop/resxtop utility— vSphere 4.0 Update 2 includes an enhancement of the performance monitoring utilities, esxtop and resxtop. The esxtop/resxtop utilities now provide visibility into the performance of NFS datastores in that they display the following statistics for NFS datastores: Reads/s, writes/s, MBreads/s,MBwrtn/s, cmds/s, GAVG/s(guest latency).

The ESX U2 release notes are here: http://www.vmware.com/support/vsphere4/doc/vsp_esx40_u2_rel_notes.html

The vCenter U2 release notes are here: http://www.vmware.com/support/vsphere4/doc/vsp_vc40_u2_rel_notes.html

.Net patch (9th June 2010) breaks vSphere Client

I’ve just been informed that Microsoft released a patch on Tuesday (9th June 2010) to update .NET which has effetely stopped any versions of the vSphere client prior to U1 working.

The only fix is to download the latest client from VMware.

The following kb article explains this. http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1022611

Enable MMU on VM’s running Citrix PS

I was recently involved with troubleshooting some performance issues running PS 4.5 inside VM’s on ESX 4.0.

When comparing the performance of a physical server to that of the VM, certain login tasks would be completed 20/30 seconds slower on the VM. We performed the usual step by step troubleshooting approach tweaking memory, network, CPU, disk placement, even running 1 VM on a host but we couldn’t reduce this time.

Then, we looked at enabling MMU on the guest. By default the setting is automatic, so I still don’t understand why it didn’t kick in, but once we forced the setting, the VM matched the original physical Citrix server on all tests.

The only this we couldn’t achieve was the same quantity of users on the VM. We went from 50 per physical server to 25 on the VM. But, we were able to run 12 virtual machine instances on a single ESX host, so actually we saw a large increase in the number of users per physical box!

The VM configuration we used was 2 x vCPU, 4GB RAM. We also enabled the vmxnet3 cards which appeared to improve performance slightly.

Using Hitachi Arrays with SRM – What you need to know

I say Hitachi, but actually both SRM solutions I’ve put together were on re-badged arrays:

•  HP XP 20000
• Sun StorageTek 9985v

The key points to make note of when deploying SRM with the above arrays are listed below:

• You must install RAID Manager on the same server as SRM
• The SRM server must be presented with a RAW disk from the array to use as the command device
• A HORCM file will be required on the SRM server in both protected & recovery sites, it should be copied to C:\Windows
• HORCM must be configured to run as a windows service
• You will need somone who is familiar with HORCM file creation to assist you!
• Hitachi SRA 2 should be installed

Once you’ve configured the above, SRM works like a charm.

SRM & Hitachi Arrays, SRA document location

If like me you’ve configured SRM for use with a Hitachi array, you will notice that the SRA doesn’t come with any documentation or readme.

Not to worry, I’ve discovered you can download a very well written document directly from HDS at the following location:

http://hds.co.uk/assets/pdf/hitachi-storage-replication-adapter-software-vmware-vcenter-site-recovery-manager-deployment-guide.pdf

vSphere design course – feedback

Last month I attended the VMware vSphere design course and thought since I’ve not posted anything since December I’d give this a quick review.

Firstly, the course is not hands on, so don’t think you’ll be implementing what you’ve learnt as this isn’t the idea.

I liked the discussion based approach and the level of expertise in the room was fairly high, so building upon the already good quality contents in the manuals made this a very enjoyable course.

Secondly I was pleased that the course materials were up to date, basing everything on ESX 4.0 U1 and talking about ESXi as the best approach moving forwards.

It was interesting that everyone on the course worked for a VMware partner. It appears the need to attend various courses to maintain a certain level of partner status is forcing people to pick up the training manuals quicker than previously. I also attended the DR & Business Continuity training towards the back end of last year which was exactly the same.

In summary, if you have a need to design VMware solutions and want to understand the best practices, this is the course for you.

Also, if you’re thinking of sitting the new VCAP design exam and already have a VCP4 this will be useful.

I’m planning to sit the design exam in August/Sept so will report back then.

Issues with ESX 4.x & HP Virtual Connect Flex-10 firmware v2.30

Anyone else seeing the issue below?

I’m using HP BL685G6 blades but I understand this issue also happens with the half heights.

Virtual Connect Flex-10 modules running firmware v2.30

ESX 4.0 or 4.0 U1, either shows the issue.

What we’re seeing is when an uplink from the Flex-10 module is pulled out, the status of the vmnics does not change. We should see the vmnics as ‘down’ but they continue to show as active.

Although an alert is shown in vCenter to say redundancy is lost, you do not see any red crosses by the cards in the networking configuration screen.

Because of this issue, unless you configure network failover to use beacon probing, you will lose connection if a link goes down.

I’ve actually posted the same content on the VMTN in an attempt to gain some feedback which has started:

http://communities.vmware.com/thread/247205?tstart=0

A call is currently open with HP and I’d expect to feedback more very soon.

B.T.W – Firmware v2.31 doesn’t make any difference

Follow up — 18th Dec 2009

It seems this is a known issue within HP, but you won’t find any kb articles or information in the public domain. 

The following gives a little more background on the issue and when we’re likely to see a fix.

Flex-10 Virtual Connect  Ethernet  modules support Smartlink per flexnic as of 2.30 Virtual Connect Manager. It uses DCC to accomplish this. (Device Control Channel: method for Virtual Connect to change Flex-10 NIC port settings on the fly without power no/off).

Prior to Virtual Connect Manager version 2.30, flex-10 Ethernet modules would shutdown all 4 flexnics, which could be a real problem if each flexnic was connected to a different network.

For example, flexnics 1a,1b,1c,1d connected to networks Red, Green, Blue, Black and network Black lost its uplink port , Smartlink would disable all 4 flexnixs. With Virtual Connect Manager version 2.30 or above Smartlink will disable only the flexnic connected to the network that triggered Smartlink. In the above example only flexnics assigned to network Black will be disabled.

Requirements:
The requirements for Virtual Connect Manager  2.30 or above to support Smartlink with DCC are:

1.  The  operation system driver must support DCC (VDB driver version 5.0.32.0 for Windows2003/2008). For ESX, VMware will have to supply the driver.

So when will VMware have a driver to support DCC?

The answer is hopefully the back end of January 2010. I understand Broadcom have written the driver and it’s currently going through testing with HP and VMware to gain certification.

Update June 8th 2010

Sorry it’s taken me so long to update this blog but the driver required to fix the issue can be located at the following location:

http://www.vmware.com/support/vsphere4/doc/drivercd/esx40-net-bnx2x_400.1.48.107-1.0.4.html

Host Profiles – Excellent, but read the label

Whilst troubleshooting my recent Platespin Migrate issues, I came across something I didn’t expect with host profiles.

Although a host profile takes the configuration of a reference host and can be used to compare against other ESX hosts, provided everything in the profile is present and correct, you’ll get a nice green tick.

What if my target server has 10 more port groups which don’t actually exist on the original reference server?

The answer is, it doesn’t make any difference, your server will be deemed compliant since it met the configuration of the reference host.

This makes sense, since everything else matches, but I’d like to see a list of differences in a future release.

I’ve tested my theory by creating an ESX server which contains 10 port groups beneath vSwitch0 and then attached a host profile from my reference host without any port groups. Running a compliance check actually shows this host as OK with a green tick.