The Last Octet

I was going through some old stuff at my parent's house the other day and unearthed my first "PC". This is not to say I found my first computer, as I had numerous computers prior to this PC. This particular machine would be the one that would lead me down the PC/x86 road and introduce me to my present day career.

I'd often wondered what state the machine would be in if I were to power it up again. I didn't have very high hopes, but I did wish that I could recover the machine's state to retain that "era" of my computing past. More on this later.

When I dug the machine out, I didn't do much prep. I plugged in the power cable and keyboard in and threw the power switch. What happened next was a scene out of my own Back to the Future movie. First there was a short pause, then a bright glowing (firey) light shown from behind the brightness and contrast dials,followed by a capacitor blowing POP!. This happened in about 2 seconds time. Then.......I saw the machine performing its memory check as it proceeded to POST. Shockingly there was very little "electronic smoke" smell in the air.

After about a minute, the machine had booted up to Norton Commander and the menu system I'd written to assist other family members who didn't know DOS.

So here I am. 20 year old computer that I want to preserve. I'm going to ponder the direction I take with this over the coming days and figure out what my approach is going to be. Those with suggestions, post'em up.

Quick cheat sheet on how to put ESXi/VMvisor on a USB stick using linux:

1. Download ESXi
2. Open the ISO (Archive Mounter works)
3. Open the image.tgz file
4. Drill down and locate the VMware-VMvisor......dd.bz2 file
5. Open the dd.bz2 file, and extract the dd file
6. Finally from your terminal, execute the following dd command using the file dd file extracted: sudo dd if=extractedfilename.dd of=/dev/USBDEVICE

On a side note, a quick look at GParted usually reminds me of the device name associated with my usb stick.

I have an old machine stockpiled full of old medium capacity hard drives that I use as my weekly backup/dump nfs destination for some virtual machine images. Thing about this particular machine is that it doesn't do anything else the rest of the week except spew heat and eat electricity.

I decided to try to use Wake On LAN to help get the machine started up when it was needed. Not knowing how to get WOL configured within Linux I set out on a Google search to help me figure out what my options were. Surprisingly, the first hit came back with what I needed. The link to the Ubuntu How To can be found here:

HOW TO: power on remotely with WakeOnLan (WOL)

The readers digest version is:

Get ethtool on the Debian backup to disk (b2d) server:

apt-get install ethtool

Edit your /etc/rc.local file on the b2d server and a add this before the exit 0

ethtool -s eth0 wol g

This configures eth0 to except the magic packet to wake up when the machine is shutdown.

If you're using a linux box somewhere on your network, you can schedule a cron job to wake up the b2d server before the backups are schedule to run. As the link states, this is simple as well.

Get the wakeonlan utility on the remote server:

apt-get install wakeonlan

Write a script that includes the following statement, and schedule a cron job to execute it 10 minutes or so before your backups are schedule to run:

wakeonlan -i B2DServersIP B2DServersMACAddr

Finally, in my situation, the backup jobs are executed from the B2D server so I included a shutdown statement at the end of the backup script (the one from the reference link works fine):

shutdown -h -P now

This makes sure the server is turned off after the backups run.

That's it. Now the B2D server runs 164 less hours a week.

Here are the symptoms and what we found resolved the matter.

We'd recently discovered that some our VM's were not successfully completing their VMware Update Manager baseline scans. Each of the problematic VM's had similar symptoms, we'd initiate the scan and then at about 55% of the way through the scan would stall for about ten minutes and then timeout.

After ensuring each of the troublesome VM's had a CD-ROM drive connected (required for VUM) we still continued to run into the stalling problem. Digging further into the issue found that the pattern was isolated to VM's that had been P2V'ed. One particular VM led us down the path to resolving the issue for the rest of the VM's. The VM in question had an exclamation mark over one of its IDE controllers in Windows device manager. Removing the conflicting device allowed VUM to properly mount the virtual CD it needed, and allowed the scan to complete properly on that machine.

The solution was similar on the remaining machines, but not as obvious. When we dug into the device manager of the other problematic machines they did not have the same Windows hardware conflicts as the first machine. Even after configuring device manager to show hidden devices, we found no offending hardware. Only after following the instructions in Microsoft KB article 315539 were we able to remove the offending legacy hardware. Those steps are basically:

1. Open the command prompt.
2. At the prompt enter: set devmgr_show_nonpresent_devices=1
3. Open device manager with: start devmgmt.msc
4. Finally, click "Show hidden devices" on the View menu

As it turns out all of the P2V'ed machines were never completely flushed of their previous Windows hardware profiles (although we'd done a regular clean up after they were migrated to remove legacy devices and OEM drivers and utilities). We also thought that we'd be able to obtain some of the problem indicators from the VUM logs, but the most we got out of those was that the scans were unable to mount the CD-ROM.

Another lesson learned.

With Update 4 for for VirtualCenter 2.5 VMware introduced a new plugin that gives a nice MRTG style graphs for host resources. The plugin adds a new tab to the host view in the VI client

Under the tab you'll find graphs for

• CPU%
• CPU MHz
• Memory %
• Disk KBps
• Network Mbps

Although this isn't a ground breaking feature, it does provide a nice "at glance" view that could be used to track down trends occurring on individual hosts. I do find it interesting that this was chosen to be added as plugin as opposed to something like a "preset" button, or drop down option to the existing performance tab.

The neatest option in the performance overview plugin is the "Virtual Machine" view. It gives a stacked graph overview of the aforementioned resources for each guest residing on the selected host.

It would be nice if there was a print or report export option associated with the overview plugin that could be used to capture the views presented. In the end, the new plugin is a nice add on that could be helpful in reviewing the performance of a particular host and the guests residing on it.