How to set-up a 2-cluster node clustered server, using Windows 2000 AS

[heatwave]

How to set-up a 2-cluster node clustered server, using Windows 2000 AS

Now I can contribute a tutorial that I made my self…sorry for the first tutorial I posted.

I won’t be explaining much on what MS clustering, there are lots of information on the web to explain what MS clustering is all about, I’m coming up on this tutorial to share my experience in setting up MS cluster. I have provided a link on what is MS clustering all about.

http://www.paraflow.bg/en/solutions/6.html


First: Checklist

1 shared storage device with at least 5 SCSI hot swappable disk drives (2 to be configured as quorum drive with RAID level 1 or disk mirroring, and 3 disks as data drive with RAID level 5 or disk stripping with parity (RAID levels can be searched over the net to give you a better understanding of it)

2 dual channel SCSI controller (one for each server and to be attached to the SSD)

(Please note that there more cheaper alternatives for your SSD solution, instead of using SCSI drives and SSDs (we all know that this is not cheap), you can also use IDE drives (ultra ata/66 or better) and a SSDs that supports Ultra-ATA drives.
*SSD = Shared Storage Device

4 NICs (2 for each PC, Why 2 for each? One will be used as the cluster heart beat, the other will be for your private network)

1 crossover UTP cable.
2 completely built PCs with 133-MHz Pentium or higher central processing unit (CPU). A maximum of eight CPUs per computer are supported.
256 megabytes (MB) of RAM recommended minimum (128 MB minimum supported; 8 gigabytes (GB) maximum).
1 free PCI slots for the SCSI controller (one for each PC)
A hard disk partition with enough free space to accommodate the setup process. The minimum amount of space required will be approximately 1 GB. More space might be needed, so it’s better to have more than the minimum requirement.
CD-ROM drive (required in installing OS, better use one for each PC to save you trouble in transferring the drive from one PC to another) FDDs will be required also since SCSI device drivers during the installation process will be searched by the installer from the FDD
2 licenses for Windows 2000 advanced $erver L
5 free IP address, 3 will be of the same subnet as your LAN (one for each PC and one as your Virtual IP or cluster IP, the remaining 2 will be the cluster heartbeat.

Second: Interconnecting your devices.

First thing for you to do is put one of your SCSI controllers in your first cluster node (lets label it as NODE1), after attaching the controller and the 2 NICs, set-up the SSD by inserting all your HDDs in it.

Attach the SSD to NODE1 (cables come together with the SCSI controller to the SSD)

Boot up NODE1 then enter the SCSI utility, this is the part where you will configure your SSD.
Configure the 2 drives with RAID level 1, this will be used as your quorum drive.

Configure the remaining 3 drives with RAID level 5, this will be used as your data drive. (This process will require you to format the configured drives and may require several reboot). Turn off everything after configuring.

On each PC, the second NICs should be interconnected via crossover cable and the first NISc on your private LAN

Installing OS

NODE1 Installation

Turn on the SSD and wait for the drives to spin and be on ready mode

Boot up the PC from the CD-ROM with the win2k AS.

Press F6 for you to install the SCSI controller, insert the driver diskette in the FDD when asked.

Continue with the installation process

Take note of the network configuration, your first NIC should be of the same subnet of your private network, this will be used by your LAN to access the clustered server and the second NIC can be of any IP as it will only be used by the cluster for heartbeat or error/fault detection (10.10.10.1 can be a good IP to set)

If you’re finished with the OS installation and properly booted up with windows, check if you can access the configured drives in your SSD, you should have multiple disks in your “my computer”.
Assuming that all is configured well, we can go on to the next node, which we will label as NODE2. But before starting with node 2 make sure to shutdown the first node and leave the SSD running

NODE2 installation

First is to configure NODE2 to recognize the SSD by following the steps in interconnecting your devices. Follow the same instructions in installing your OS, first NIC should be configured to have the same IP series of your LAN and the 2nd NIC will be for the heartbeat IP again (10.10.10.2 same subnet as what we configured for NODE1) If your done with the installation, shutdown everything.



Third: Configuring MSCS

Take note of your Domain Controller’s IP (if you don’t have a DC, you can set up your first node with Active Directory and configure it as your DC, although using your cluster node as the DC is not a good practice, I’m just suggesting it to simplify things. It is recommended that you use a separate DC for your network.

NOTE: Never forget to add the DC’s IP on the network configuration of your 2 PCs network property as the DNS IP or else you won’t be able to join the domain.

For the heartbeat NICs you will need to tick the disable net bios over TCP from the WINS tab of the network properties.

Boot up SSD then followed by NODE1

Go to start, settings, control panel, administrative tools, configure your server, then choose active directory, you will have to set this up to join your domain.

After the AD set-up, go to star, settings control panel, administrative tools, advance, and then select cluster installation.

Just follow the step-by-step instruction in setting up the cluster, you will need to make some adjustments based on your requirements/configuration.




Well, that’s it, enjoy using your cluster and explore the wonder of a high available set-up. I hope you all find this useful enough.

[Deaflamb]

Nice Post. Havn't read much about Microsoft Clustering Technology. Here is a link for setting up a Beowulf Cluster . I know that they are diffrent types of clusters, the Microsoft as I understand it being a backup type system and the Beowulf Cluster being a parallel processing cluster. But they are both interesting so I though some might enjoy the reading.

Nice Post

DeafLamb

[mohaughn]

A couple of corrections

1) You only enter the GC as the DNS server is you are running DNS on your GC. I would actually recommend against running DNS on your GC.

2) Not only should you disable NLB over the cluster/heartbeat nic, but you should also disable automatic DNS registration. If you don't you will get DNS client errors in your system log because this NIC can't properly register itself in DNS.

3) I would recommend installing the OS on both systems before you worry about building the cluster. It is must easier to get both systems online and configured, and then install clustering on both nodes. Just make sure that you do not install your cluster aware applications until after you install clustering.


Some things you should never do with an MS cluster-

1) Have both nodes up with the cluster service stopped on both nodes. This will cause your shared disks to get corrupted. Basically the two machines fight over who should have access to the disks.

2) Restart both machines at the same time. If the cluster service starts on both machines at the same time you run the risk of corrupting your external drives as they will fight for ownership. You should always start one machine, make sure it is up and running, and then start the other machine.

3) Stop services through the control panel. If you want to start/stop a cluster resource, always always start it through the cluster administrator. Otherwise clustering is not aware that the administrator requested the resource to stop and it will treat the service stopping as a failure and take whatever actions you have it configured to take.

[bballad]

mohaughn: you can minimize those curroption issues by makeing sure you have every thing setup as a failover cluster basicly one server is the main server and if it goes down the other takes over, no fighting for disks...running a SQL cluster this way.

[heatwave]

[i] A couple of corrections

1) You only enter the GC as the DNS server is you are running DNS on your GC. I would actually recommend against running DNS on your GC.

"Take note of your Domain Controller’s IP (if you don’t have a DC, you can set up your first node with Active Directory and configure it as your DC, although using your cluster node as the DC is not a good practice, I’m just suggesting it to simplify things. It is recommended that you use a separate DC for your network"

[i]
2) Not only should you disable NLB over the cluster/heartbeat nic, but you should also disable automatic DNS registration. If you don't you will get DNS client errors in your system log because this NIC can't properly register itself in DNS.[/B]

Thanks for the additional infos.

[i]
3) I would recommend installing the OS on both systems before you worry about building the cluster. It is must easier to get both systems online and configured, and then install clustering on both nodes. Just make sure that you do not install your cluster aware applications until after you install clustering.[/B]

I think i've mentioned something about having the OS installed first.

[i]Some things you should never do with an MS cluster-

1) Have both nodes up with the cluster service stopped on both nodes. This will cause your shared disks to get corrupted. Basically the two machines fight over who should have access to the disks.

2) Restart both machines at the same time. If the cluster service starts on both machines at the same time you run the risk of corrupting your external drives as they will fight for ownership. You should always start one machine, make sure it is up and running, and then start the other machine.

3) Stop services through the control panel. If you want to start/stop a cluster resource, always always start it through the cluster administrator. Otherwise clustering is not aware that the administrator requested the resource to stop and it will treat the service stopping as a failure and take whatever actions you have it configured to take. [/B]

Thanks for the corrections and for the additional infos, it's a big help if others will give some corrections and additional info as it will make the informations more accurate. thanks again.

[mohaughn]

Originally posted here by bballad
mohaughn: you can minimize those curroption issues by makeing sure you have every thing setup as a failover cluster basicly one server is the main server and if it goes down the other takes over, no fighting for disks...running a SQL cluster this way.

bbalad- I think a lot of the corruption issues have been fixed with the later service packs for Win2k. There were a lot of corruption issues running on NT4.0, and when Win2k came out we actually reported several bugs to MS in regards to disk corruption. We run all of our servers in active passive using a Compaq Storage Works San. The secure path software that handles the pathing for the fiber optic SAN has also improved dramatically to lessen to chances of the nodes fighting for disk ownership.

However, stopping the cluster service on both nodes is still a no-no, unless you know for sure that your hardware configuration will prevent disk corruption. The current compaq/hp high end systems don't even prevent corruption in this particular instance.

If you do experience disk corruption, here is the easiest way to repair the quorom drive.

http://support.microsoft.com/default...b;en-us;257905



Here are a few of the problems that I have run across running MS cluster. If you are not running the latest service pack, you should be aware of the problems.

http://support.microsoft.com/default...b;en-us;295300
http://support.microsoft.com/default...b;en-us;307939
http://support.microsoft.com/default...b;en-us;293778
http://support.microsoft.com/default...b;en-us;815616

[maxim_86ualb2]

nice ....... love it