VIOPS: Proven Practices for Deploying and Managing VMware: Message List

Re: HP Blades, Virtual Connect, 10Gb, Flex-10 and vSphere

dconvery — Thu, 10 Sep 2009 00:32:02 GMT

Thanks for clarifying! I wasn't so sure at the time. Anyway, I did some additional research and found these:

So-so video -> http://vimeo.com/6300062

Decent preso -> https://www.myciscocommunity.com/docs/DOC-9686

Re: HP Blades, Virtual Connect, 10Gb, Flex-10 and vSphere

dconvery — Mon, 07 Sep 2009 15:11:51 GMT

I know you can create virtual NICs to regulate bandwidth, but that may prohibit using VLANs

Re: HP Blades, Virtual Connect, 10Gb, Flex-10 and vSphere

dconvery — Fri, 21 Aug 2009 12:55:56 GMT

I am going to attempt to answer at leats some of your questions. These answers are really just my opinion and I am sure someone will argue different configurations:

Right now, a single 10GbE uplink may be OK for you. I would recommend at least two per I/O module if you can afford it. This is more about redundancy than bandwidth.The best bet is to measure networking performance of the VMs or P2V candidates to see if you have a risk of saturating theuplink. Don't forget the NFS I/O as well. If you expect higher I/O, certainly max out youor uplink connections now while your are in the planning stage. (How's that for wishy washy?)

Stacking the chassis simplifies management. You should also spread your ESX cluster across at leats two chassis. Check out Duncan Epping's HA deep dive -> http://www.yellow-bricks.com/vmware-high-availability-deepdiv/

With 10GbE, you can share uplinks if you are segregating traffic via VLANs. If you have a separate physical network for NFS, then obviously you need separate uplinks.

I usually recommend passing the VLANs through, especially if you are planning to use the 1000v dvSwitches. This allows you pass the VLANs up to the dvSwitch and, with the 1000v, you can manage policies, security, etc. right up to the VM.

Let VMware HA and FT do their jobs, keep it simple when it comes to fail over.

Using 10GbE as one single link is OK, but see below for some different takes when thinking about your NAS connections. Leaving a single channel allows the bandwidth to be used as needed for your applications (workloads), management, VMotion and NFS/ iSCSI.

Separating your NFS traffic is recommended. In my opinion, VLANs are fine for this. You should also have separation of management traffic (between vCenter and ESX) and VMotion traffic as well. So, at least four VLANs: Management, VMotion, NFS, and VM traffic. You should use PortGroups to separate VLANs at the dvSwitch if you are not using 1000v. Check out Ken Cline's excellent post on Virtual Networking -> http://kensvirtualreality.wordpress.com/2009/03/29/the-great-vswitch-debate-part -1/

I like the idea of more, less expensive DIMMs, so go with the BL490c. My opinion is to use embedded ESXi. The USB stick is a SPOF, but so is an internal SSD and VMware HA will cover this for you. And, it you set up HP SIM properly, you can do a predictive failure and make it graceful.

For a good reference architecture, check out http://www.hp.com/go/matrix

Dave Convery

VMware vExpert 2009

http://www.dailyhypervisor.com

Re: vSphere in a box - Thoughts & suggestions.

dconvery — Tue, 11 Aug 2009 15:26:24 GMT

Just to follow up and "complete" the information:

Host system must have Intel-VT / AMD-V enabled in the BIOS. Many systems do NOT have this enabled by default.

This configuation does not seem to be "supported" by VMware at this time, but community support is always available.

Re: Unkown VCB SSPI Authentication Problem

dconvery — Fri, 17 Jul 2009 15:05:31 GMT

Try flipping up to verbose logging by adding the "-L 6" option. So your command would look like this:

"E:\vcbMounter.exe -h virtualcenterservername -u __sspi__ -p "" -a name:virtualservername -r E:\virtualservername-%DATE:/=% -t fullvm -m san -L 6"

Copy and paste the command line output to a text file and attach the file to this thread. Also attach the logs in %TEMP%.

Re: Partition sizing via Kickstart script

dconvery — Fri, 29 May 2009 13:49:44 GMT

I think this has to do with the geometry of the disks provided. I believe it is a function of the RAID adapter a little bit as well. If you do a DF -h what does that report?

Re: VCB Backup solution

dconvery — Tue, 26 May 2009 15:07:44 GMT

Thomas -

Thanks for re-posting this to VI:OPS! Since this has turned into a guide for a working VCB script, where better than posting it here.

For those of you just reading this post, it is a contiuation of a blog post on the VMTN site -> http://communities.vmware.com/blogs/jturver/2008/01/24/vcb-how-to-backup-a-selec ted-specific-disk-or-disks-using-vcbexport-and-a-useful-script-to-do-it

Dave

Re: vSphere in a box - Thoughts & suggestions.

dconvery — Thu, 21 May 2009 11:58:31 GMT

Specs:

Lenovo ThinkPad T61p

Intel(R) Core(TM)2 Duo CPU T7500 @ 2.20GHz

4GB RAM

Host OS = Ubuntu 9.04 running VMware Workstation 6.5.2

The ESX or ESXi Guest is allocated 2GB RAM

The vCenter VM is allocated 512MB RAM, which is OK with SQLExpress and a small vCenter.

The W2K3 VM inside the ESX VM is only allocated about 128-256MB RAM and it runs slowly, but it is great for testing.

I have used this setup six ESX 3.0.x for testing things like VI Toolkit/PowersHell/ PowerCLI scripts, VCB scenarios, DRS testing, etc.

If you look at these sites, you will see some more info on the processes.: http://www.rtfm-ed.co.uk/ and http://xtravirt.com/

Incidentally, I heard a rumour that VMware made an effort to make sure we could create an ESX/ESXi VM inside of an ESX/ESXi host for testing and lab purposes. It just so happens that it works on WS as well.

Dave

Re: vSphere in a box - Thoughts & suggestions.

dconvery — Wed, 20 May 2009 13:11:33 GMT

Darren -

Check out my recent post about running VI4 inside of VMware Workstation -> http://www.dailyhypervisor.com/2009/04/29/running-vmware-esx-4-rc-in-a-vmware-65 2-vm/

It really just lays out the vmx settings.

Dave

Re: How do you do VI3 Log Management?

dconvery — Tue, 31 Mar 2009 13:20:35 GMT

Rich -

You CAN set up a relay to a syslog server on ESXi.You set it up in the advanced settings or via powershell and the VI Toolkit.

There are two big reasons for syslog servers, the first is to gather logs for security reasons so that we have anuntouched set of logs in case of intrusion or corruption. The second is to have a centralized repository of all of the log information.

I haven't done a huge amount of research yet, but I found so far that there are several alternatives for ESX/ESXi syslog collection. One is syslogd and phplogcon another is Kiwi and a third is Splunk. I am sure there are many more, but I also was looking for things that could aggregate the logs into a timeline for troubleshooting issues and do it in a straight-forward manner.

So far, I like Splunk for one huge reason: It can use WMI connections to Windows servers to gather event data. That gives Splunk the ability to provide a syslog service to all platforms in a data center, not just unix or not just VMware.

Dave