philipnet

I received the error message "An error occurred while accessing the DHCP database. Look at the DHCP server event log for more information on this error" when administering DHCP this afternoon.

It took me ages to track down what the problem was - the clue was that I was adding a DHCP reservation; every time I clicked on "Add" I got the error message.

To troubleshoot it I wanted to get an export or dump of all the entries and configuration of DHCP on the server. After a bit of Googling I came across this:

netsh dhcp server <IP address> dump >u:\dhcpdump.txt

But the first time I ran the command I got the output

The following command was not found: dhcp server <IP address> dump

The answer is to issue the command:

netsh add helper dhcpmon.dll

You just have to do it once - and then netsh will know about the DHCP commands you can use.

N.B. <IP address> has to be the IP address of the DHCP server. It can't lookup the server from the hostname

As for my problem; searching for the MAC address I was trying to use highlighted the issue. A single MAC address cannot be reserved twice in a scope. Unfortunately the error message I was getting wasn't explicit enough (and I don't have the rights to look at the logs on the DCs).

I’ve been able to isolate my VMWare BSODs to hard drive issues.

No hard drive has failed – yet, but my first clue was when the Windows crash reports stopped saying “Unknown error” and started saying “Hard drive problem – probably nothing to worry about”  .

That “nothing to worry about” is actually:
ATA Error Count: 38

Which I found out by logging on to the VMWare host and issuing the commands:
smartctl --all /dev/sda
smartctl --all /dev/sdb
as the VMWare installation is RAID 1-ed across both drives

As there isn’t anything wrong with sdb I can safely switch sda for a new hard drive and rebuild the RAID, but that means a few days of little activity on the system.

Is it me, or are you getting BSODs when installing Server 2003 SP2?

I suspect that it’s the VMWare environment that I’m running it under that’s causing the odd problem – how can I get a BSOD under a 2GB 2processor VMWare machine when I don’t get one with a 512MB 1processor VMWare machine?

Host memory?
Host swap?
Host load?
Host hardware?

On the bright note I did discover that if the service pack fails to install and the system is still bootable, the service pack will uninstall itself.

Through experimentation I found that pre-allocating VMWare disks is the best course of action. Sure, you might over allocate disks sizes on the host, but that can save having to restore a from backup when things go wrong.

I recent had the opportunity to experiment with allocating disk or not allocating disks under VMWare. I was under the impression that pre-allocating VMWare disks was the preferred option, but I wanted to know why. But I found out at my cost.

I wanted to build another DC on my test network. A reasonably high spec PC running VMWare was to be the host and a low 512MB RAM, 10GB and SCSI HDD was to be the VMWare machine.

I installed Window Server 2003 Standard Edition, set it up, added it to the domain and was set to go. I thought that installing SP2 would be prudent before making it a DC; and so I set that going.

Unfortunately half hour later, I was greeted with a BSOD and a rather lengthy memory dump. So I reset the VMWare machine and tried to repair Windows. It wasn’t having any of it, and so half hour later I gave up and did a complete re-install.

Thinking something just went wrong with the first attempt, I updated the drivers before installing SP2 for the second time. The drivers were up to date, so I installed SP2, and thirty minutes latter was greeted by another BSOD.

I was getting rather fed up of this. So I deleted the VMWare hard drive and created a new one – all allocated – for the third attempt.

This time it worked. No BSODs and now one happy DC!

Something I’ve been looking for is the ability to link my WindowsAdvice.com blog with my FaceBook profile.

Now, with thanks to this post by Josh Ledgard, I can do just that!

I was looking for a few documents on VMWare performance.

I’ve recently had the notion of building myself a meaty server running Linux (Ubuntu) on which I can run test servers/desktops as much as I want. It saves running them on multiple 500MHz PIIIs which I can get cheap on eBay .

I wanted to confirm something I read on how to setup VMWare on Linux (Ubuntu) which was to configure the Guest OS with a SCSI hard drive rather than an IDE one.

So I went search for VMWare Performance docs; and found a few useful URLs:

1. Performance Tuning Best Practices for ESX Server 3 http://www.vmware.com/pdf/vi_performance_tuning.pdf
   Which details various things to ensure to get the best performance out of your VMWare server and the Guest OSes.
2. Storage Subsystem Performance in VMware ESX Server: BusLogic Versus LSI Logic http://www.vmware.com/pdf/ESX2_Storage_Performance.pdf
   Which details the speed comparison of two SCSI drivers(!)

So I’m no where near where I wanted to be. So let’s try the VMWare site; and I find the GSX manual http://www.vmware.com/pdf/gsx32vm_manual.pdf which tells me:

“The wizard recommends whether to use SCSI or IDE, based on the guest operating system installed in the virtual machine”

Some help that is! I want the answer now!

So back to Google it is and on the VMWare site I find the text:

“The maximum size is 128GB for an IDE virtual disk or 256GB for a SCSI”

So I’m getting closer. And then I find this post (and I’m about to give up!):

IDE vs SCSI under Virtual Server http://blogs.msdn.com/virtual_pc_guy/archive/2006/02/06/525487.aspx

“The final thing to consider is performance - and this is a bit tricky.  Contrary to common sense, the performance of our emulated SCSI controller is slower than that of our emulated IDE controller.  The reason for this is that the SCSI controller is a lot more complicated to emulate than the IDE controller.  Now - this changes once you have Virtual Machine Additions installed.  As part of Virtual Machine Additions we install an accelerated SCSI driver.  Once this driver is installed the performance of our emulated SCSI controller is significantly faster than our emulated IDE controller.”

Okay, so it’s the wrong software, but right camp. So one final Google later I find this forum posting:

Choosing SCSI vs. IDE for boot drive 
http://www.vmware.com/community/thread.jspa?messageID=83153

“After I failed to find DOS drivers that would let my Ghost boot disk see the SCSI boot drive, I recreated the VM with an IDE drive and all my problems went away.”
…
“Do you plan to migrate the VM to ESX at any time? If so, then you should use SCSI since ESX does not support IDE disks.”
…
“If your guest OS includes "more intelligent" SCSI drivers, then you may get better performance from a SCSI virtual disk than from an IDE virtual disk.”

So; concrete answer may be none, but the answer is there. Go IDE if you want a quite life, or go SCSI and install the VMWare Tools CD.

I’ve changed jobs and am now working for a large insurance firm in Surrey, UK.

I am firmly in second line support with delusions of 3^rd line (although those delusions are slowly turning into reality).

It’s a great environment, a great team to be in and a great computing department to be part of, work in, and work with the people there.

One of my bugbears about WSUS 2 is that it isn’t easy to find and destroy superseded updates.

The only way I’ve found to do this is to view all updates with any approval and synchronised at any time.

Then – manually – go down the huge(!) list of updates and where any update has an icon to the left (whether it’s a supersedes icon or a WSUS icon) select it, and check to see if the update is actually superseded.

I ought to put a plug in for the little utility GetRight.

I found it many many years ago and has saved me thousands of minutes (if not hours) in re-downloading files, resuming files, busting through tight bandwidth restrictions and connection ratios.

I used it today to download the latest JRE from Sun, something that IE failed to download due to a “connection timeout”  .

What’s even better is that the license I bought back in 1999 (!) is still valid for the latest version 6; although I did need to request an updated code for it to validate.

It’s been a while since I played with (or even used  ) Admin Studio SMS Edition.

It seems that MS/Macrovision have updated it beyond the free SMS version 6 they were offering, to version 8.

It can be downloaded from: http://www.macrovision.com/products/adminstudio/adminstudio/editions/smsdownload.shtml . I’ve got no idea what’s been added in the intervening years, but I’m about to find out .

I’ve downloaded the recent ITMU update for SMS 2003 that’s been mentioned at various places.

One first annoyance is that I had to uninstall my current ITMU through Add/Remove Programs. Otherwise the installation would fail with ‘an error’ – but I couldn’t find anything that told me what error!

So my immediate problem is (re-)creating the packages and collections that were deleted, or just reconfiguring them; like letting the Microsoft Updates Tool Sync run “whether or not a user is logged in” .

The next puzzlement is the Windows Update Agent. I’m going to have to read the documentation on that one!

And lastly I have to ensure that I am getting all vulnerabilities detected.

I had a DC that ran on a old PC. Whilst there was nothing wrong with the hardware, the hard drive used was 10Gb big and it made so much noise it was unbelievable!

I quickly grew tired of it – I know PCs aren’t the quietest of things but this was taking the mickey. So I hatched a plan to migrate the whole DC onto a VMWare instance on another machine.

Step 1: Install VMWare and get the VMWare console working.
I really need to mention that the destination machine was a headless Linux machine. I couldn’t get authentication across the network working, so I had to install various X and GTK+ libraries and Exceed on a client PC to complete this step.

Step 2: Migrate current HDD from old DC to new DC and create a suitable VMWare instance.
This was pretty easy. I enabled the least amount of hardware for the VMWare instance as I wanted to keep the Windows instance as light as possible.

Step 3: Migrate DC HDD to a virtual HDD.
I used PowerQuest DeployCentre to copy the contents of the old DC HDD to the new virtual one; this was within the VMWare instance and took almost an hour. Then reconfigure VMWare for only one HDD, the new virtual one.

I had completed the migration and had a really silent DC.

It’s been sometime since I posted here, or even ‘resolved’ the RPC over HTTP problems I was having.

My colleague was also looking a getting RPC over HTTP working. He had found a few posts online that stated that you need to reboot the DC in order for the changes to take effect.

Unfortunately at the time we didn’t have enough proof that that was the solution, nor was there a planned or unplanned reboot of the server; so we had to let that lapse.

A couple of months then passed and to my surprise I found that RPC over HTTP was indeed working!

I can only surmise that some sort of roll-over or re-sync occurred on our main DC or within Active Directory to enable RPC over HTTP to work.

Since then, I’ve customised the Outlook profile to use RPC over HTTP by default and have been deploying the new configuration to users on a ad-hoc basis. It’s proved immensely beneficial!

I’ve been struggling trying to get RPC over HTTP working from our main Exchange 2003 server. The box itself is on Server 2003 SP1 and is running SP1 of Exchange as well. The problem is that all RPC tests fail and Outlook constantly talks to our Windows 2000 DC regardless of whatever settings I plug into Outlook, Exchange and IIS.

Last week I hit upon a few interesting posts – it seems that some other administrators were also failing to get this to work on other Windows 2003 SP1 boxes – a factor that they seemed to think was important.

Knowing that downgrading our production email server was out of the question, I’ve started building a test Exchange Front End server running Windows 2003 with only hotfix 831464 installed. I’ve installed Exchange, upgrade to SP1 and enabled Outlook Web Access (webmail).

Except when I https://webmail-test/exchange/ . I get told that “The page must be viewed over a secure channel” and to “Type https:// at the beginning of the address you are attempting to reach and press ENTER.” !

Well the rain held off until I got home – so so much for that reducing the number of people who actually attended. However there were noticeable spaces, so MS were right to overbook it (as is the general practise anyway).

So what did I get out of “The Best of MMS”?

I’ve got to say that those 15 min breaks weren’t long enough.
There were some interesting vendors there – Quest have a lot for you if you want Linux or UNIX integration with SMS and MOM and 1E were there with their SMS and MOM products as well. But those were the only two of interest that I managed to speak to.

The talks were actually pretty good. It’s a shame that the presentations did not match the printouts, particularly Kirill’s . For those who attended, you should be able to download the updated presentations from: http://www.microsoft.com/uk/windowsserversystem/mms/default.mspx 

If the printouts were any indication of the state of things at MMS 2005, then with seeing the difference I can say things are progressing along nicely. Kirill’s presentation mentioned that Server 2003 R2 is expect in the next few months with:

• More UNIX integration (a DC can be a NIS master and “Services for UNIX” is included).
• MMC 3.0 (previously referred to as MMC 2.1) will be included (and will be a downloadable update for XP SP2 and Server '03 and later) and provide a much more MOM 2005 like interface.
• WS-Management with the beginnings of integrating with the Systems Definition Model.

Vlad and Bill's presentations on MOM, SMS and WSUS/WS-Management/MMC were very informative, especially some of the walk-throughs/demos. I see why the try-out-the-software sessions were so heavily recommended for MMS.

Bill gave us some interesting definitions:
Career Limiting Move = Rebooting Bill Gate's laptop during a board meeting (as SMS had just installed some updates).
Toxic = Using OSDFP to deploy XP SP2 to 100,000 computers in one fell swoop.
Five Feet From Termination = Standing next to Bill Gates or Steve Ballmer on stage saying "Yes, Bill/Steve, click on that".

Oh, and I did manage to get an answer as to why only WSUS can deploy updated drivers. The way the current SMS Client works is by downloading the update catalogue from the a distribution point. The compressed catalogue size is about 1Gb! What Bill and the rest of the team are working at for SMS v4 is for the clients to request the updates they require.

However the steal of the show was Michael Emanuel’s talk at the end on DSI (Dynamic Systems Initiative). What he went through; where he was placing the next 10-20 years of computing was taking the consultation out of consultancy. It wasn’t just about empowering the end user, but providing inbuilt capacity planning and resource optimisation. He didn’t mention self-healing systems; but that surely has to be a part of it.

Quite frankly; expect sys-admin jobs and consultancy to change. I know I do, and that means that I’m going to have to find a new career path!