Yellow Bricks

by Duncan Epping

ESX

Distributed Power Management glitch

· ·

I was just testing the new, but still experimental feature, Distributed Power Management. I’ve installed 3 ESX Hosts and added all of them into an HA-DRS Cluster. I enabled DPM and set the mode to manual and pressed “Generate recommendations”. I had one host with an HA failure and DPM recommend me to switch off another host. Which in this case would leave me with no host for redundancy… weird.

I’ve also witnessed that the host with two VM’s running on it was recommended for a power down, instead of the host that did not have any VM’s running.

The VMware website states:

 When resource requirements of workloads increase, DPM brings powered-down hosts back online to ensure service levels are met.

I guess having a full working HA cluster is definitely top priority to meet service level agreements, so be careful with this option and test it really well…

Checking for snapshots, html email report!

· ·

When reading the VMTN forum I stumbled upon a topic about checking the VMFS for snapshots. Besides snaphunter by Xtravirt, which is an excellent script, there isn’t much out there. You can check the VirtualCenter database for snapshots but this way you would not notice the orphaned snapshots. (Orphaned snapshot occur when the vmsd file gets corrupted.) A couple of months before Xtravirt posted their script I wrote my own snapshot checking script named snapcheck.sh. [Read more…] about Checking for snapshots, html email report!

Undocumented VCB config.js feature

· ·

One of my customers wanted to use the default VCB framework but did not want to quiesce the VM for several reasons. (Databases, Active Directory etc.) I could not find an option in the config.js file but noticed the following in the file glue.js:

// A fallback to be able to switch to non-quiesced snapshots
if (typeof(NO_QUIESCE) != "undefined") {
cmd +="-Q 0 ";
}

In other words, setting the option “NO_QUIESCE” with no value in config.js results in the VM not being quiesced, default it will quiesce the VM! I added the following line to the  config.js file to accomplish this:

NO_QUIESCE="";

P2V’ing a Suse Linux server

· ·

I just P2V’ed a Suse Linux Server with the cold boot iso. When booting I was confronted with a kernel panic. This is caused by the fact that the modules for the LSI Logic or Bus Logic SCSI card aren’t available, which causes the root volume to be unmountable. I tried to boot from the installation CD with the “rescue” option but this did not provide me with a fast solution for the problem. But the following was a quick fix:

  1. Boot from Suse CD
  2. Select to setup a new system
  3. Select to boot from the currently installed OS
  4. Run Yast and configure the correct SCSI controller
  5. Remove all the NICS and configure a new NIC
  6. Install VMware tools

Swap files in 3.5 and the EMC Celerra PDF

· ·

The PDF I blogged about yesterday included info about 3.0.x and 3.5 features. One of the things that was missing was the option to have the swapfiles created locally. In 3.0.x this done by setting the option “sched.swap.dir” in the .vmx file. The only problem is that this prohibits you from using VMotion for these VM’s. In 3.5 VMware resolved this issue. It’s now possible to set the swap file location to local storage and keep VMotion at the same time. This can be done very easily the following way:

  1. Edit setting on cluster, click “Swapfile location” and select “Store the swapfile in the datastore specified by the host”

  2. Click on the ESX Host, click on the “Configuration” tab, click on “Virtual Machine Swapfile Location” and “edit”
    Select “Store the swapfile in a swapfile datastore selected below”
    Select the local attached storage: [hostname:storage1]