BDD: Revenge of nVidia

OK so maybe this one isnt nVidias fault but its always nice to point the finger at them after all that BSOD in Vista due to crappy drivers.....thats another story....

Having had the joy of booting into the BDD image and getting the NIC to see my distribution share we moved on to getting the image installed, no problem until after the first boot and I get a 'disk read error has occurred' which appears to come straight after the BIOS.

First up, its not likely your drive if you are in the middle of an install so dont panic. Secondly there are some fixes (http://support.microsoft.com/kb/931760):

In short:

1) Change the BIOS settings from your drive to LARGE from AUTO, it will still be full size.
2) Get hotfix 931760, this can be downloaded from the nice chaps at http://thehotfixshare.net/

The problem has something to do with the XP kernel hence a hotfix for XP, fixed in SP3 apparently but who wants to jump the gun early on that one eh?

BDD 2007 and nForce

I just started experimenting with BDD 2007 with MDT 2008, great product and the right price - FREE!

Got the whole thing setup, smacked in some drivers, XP with SP2, few apps for testing, under an older test computer with an nForce2 mobo it worked a treat except for a few app tweaks but under the newer Shuttle SN68SG2 systems we use in house there is another issue.

The Win PE 2 that is used is based on Vista so Vista drivers are required for it to build the PE image that boots from the network, fine no probs, however the driver used a two-tier model as discussed here: http://www.deployvista.com/Blog/JohanArwidmark/tabid/78/EntryID/29/language/en-US/Default.aspx

Now this is all well and good however it doesnt work! Looking at other drivers I have added for the nForce2 mobo the two-tier driver did as it was told and a 'Network Bus Enumerator' driver is added as well automatically. With the Vista on the 630a drivers this doesn't happen so when I boot my image it simply doesnt work, although it boots off the network image entering the user details for the share simply says 'Network Not Found', running IPConfig shows only the internal network adapter :(

Lets hope nVidia can sort out there drivers once and for all, since the days of playing with BartPE nForce drivers have never been easy.

UPDATE: It would seem that the Vista driver for this type of board works differently, I removed all the nVidia drivers from BDD and updated the install point and readded the WIM image, booting now works.....on to the SATA driver now :(

FlexLM and ESX

We were a little worried about migrating a server running FlexLM as the licensing was tied to the MAC address of the network card and we had never tried this before.

We took the plunge however with a couple of potential fixes in the pipeline. Well first fix and a winner! All we had to do on the guest was enter the MAC address of the old server into the advanced settings of the network card configuration. Windows did the rest and as it is in the VM itself this means VMotion will work too.

The second fix involved editing config files on the VM to hardcode the MAC onto the ESX servers virtual cards, this of course would mean the server would have to reside on that virtual card at all times which could be a problem when using HA and VMotion.

Windows Time and VMWare - solution?

I think I may have resolved our wandering time problems. I finally attacked the ESX servers. Logging into the console allow the ntpclient to communicate:
esxcfg-firewall --enableService ntpclient
Than make sure it starts after any restarts
chkconfig ntpd on
Configure the NTP service, http://www.tlanews.com/TLA/NEWS/2006vmware/2007-02-22-ntp-esx.htm helped me get to grips with this. I used the time server pool at pool.ntp.org for my list, also check you add the nomodify line from the site as this restricts abuse.
Once your list is in and you have also updated the step-tickers file with your server list you can restart the NTP service:
service ntpd restart
This should get the system a new time, you can then set this to the hardware clock:
hwclock --systohc
You now have a correct hardware time and system time that is kept up to date via the external servers.

Now this is where is gets complex, different people say different things including VMWare and Microsoft so this is my solution and so far so good. Any server thats wasn't a DC had its Windows Time service stopped and disabled, then enabled the Time Sync option in the VMWare tools tray icon.

For the DC's I stopped the Windows Time service on them all except the PDC then:
w32tm /unregister
To remove all old Windows Time meddling from the past, then readd:
w32tm /register
Start up the service and:
w32tm /resync
To allow the Windows computer to get the time from its normal AD method (the PDC).

Now on the PDC edit the registry as follows under:
HKLM/System/CurrentControlSet/Services/W32time
Do the following:
Parameters\Type = "NTP"
Parameters\NtpServer = "0.pool.ntp.org 1.pool.ntp.org 2.pool.ntp.org"
Config\AnnounceFlags = "5"
TimeProviders\NtpServer\Enabled = "1"
TimeProviders\NtpClient\SpecialPollInterval = "60"

This enables it as a client and a server. Run w32tm /resync to get the new data from the new NTP server list.

OK so this gives us our usual DCs talking to the PDC as normal BUT in the event that something goes tits up the hardware clocks that they have reverted to in the past are also using the same time as the NTP service so in theory they won't change more than a quarter of a second (based on my testing of the ntp pool servers). It also means we don't have to worry about any other servers other than the DC's as they are all using the ESX server which is using the same time pool as the DC's.

Will this work? We shall wait and see.....

ESX Server Keyboard Layout AFTER Install

I had to go to the server today as part of my change in the NTP system. Lo and behold someone set it for US layout which meant all my shift + nuber keys all gave me garbage.

Here is how to fix it immediately:
loadkeys uk
And to change it to fixed edit:
/etc/sysconfig/keyboard
to show uk instead of us in my case.

w32time and VMWare

OK so VMWare is a bitch for time, due to the machines being virtual time isn't 'normal' so you have to let ESX be the host and disable the windows time service on each of the VM's and let ESX pass the time in the VMWare tools on each VM. Easy when you know how.....anyway thats another story.

How we got here is the problem, is you start setting up your VM's without doing the above you can end up with all sorts of issues where the DC VM's change time and this then makes the DC an untrusted partner for replication, thus splitting your network and allowing some clients to log on one part and some on another, in short, ugly!

A fix, 2 reg tweaks:

HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\NTDS\Parameters
ValueName = Strict Replication Consistency
Data Type = Reg_DWORD
Value Data = 0

HKEY_LOCAL_MACHINE\System\CurrentControlSet\Services\NTDS\Parameters
ValueName = Allow Replication with Divergent and Corrupt Partner
Data Type = Reg_DWORD
Value Data = 1

This allows your server that is being thought of as untrusted to do a sync as normal, once that occurs these changes can be revered, change the Strict Replication Consistency back to 1 and delete the Allow Replication.... entry.

In the meantime I am off to play with ESX firewall to set it up as an NTP server client for my remaining physical server.

Enable Remote Desktop Remotely

Nice handy one this...

1) Open regedit
2) Click CONNECT NETWORK REGISTRY from the FILE menu
3) Browse HKLM\SYSTEM\CURRENTCONTROLSET\CONTROL\TERMINAL SERVER
4) Make sure key fDenyTSConnections is set to 0
5) Add the users to the REMOTE DESKTOP USERS group on the remote PC.

Saves your legs!