The Mole 33: Technical Answers from Inside Microsoft - PDC on LAN, Raw Drives, BDCs, Stored Procedures, TCP/IP, BackTalk

  • Article

April 24, 2000

Editors Note The questions and answers below are from the Inside Microsoft column that appears regularly on the TechNet Web site at the following location: https://www.microsoft.com/technet/community/columns/insider/default.mspx. To find out how to submit questions of your own, see the end of this article or go to https://www.microsoft.com/technet/community/columns/insider/default.mspx.

The TechNet Mole provides expert answers from deep within Microsoft to questions from IT professionals. This installment focuses on these issues:

  • Mole: Genius

  • Converting Raw Drives

  • Just how Many Backup Domain Controllers Are Enough?

  • "SQL Server 7.0 Stored Procedures Give Bill a Headache"

  • TCP/IP Subnetting in 25 Words or Less

  • BackTalk!

On This Page

Mole: Genius

"Converting Raw Drives???"

Just how Many Backup Domain Controllers Are Enough?

"SQL Server 7.0 Stored Procedures Give Bill a Headache"

TCP/IP Subnetting in 25 Words or Less

BackTalk

Mole: Genius

Hello Sir,

As you are Genius, I would like to ask you one question. Here's the problem:

We have One PDC and Three BDCs in our LAN. All are functioning well till last week. Right now what is happening is when all the Domain Controllers are on, i.e. PDC and BDCs , sometimes there is no PDC on the LAN. All the Domain Controllers are BDC. Again, after some time the original PDC is there without doing anything at all. The servers are NT Server 4.0 with Service Pack 5.

What we did: We promoted one of the BDC to PDC then original PDC is automatically demoted to BDC. But what we saw again after some time was that there is not any PDC in our LAN. All Domain Controllers are BDC only.

Sir, If you know why this is happening or please give me direction how to go about it.

Santosh

Golly Santosh,

You have actually caused Mole to blush. Genius? Well, Mole wouldn't go that far (brilliant, yes) as he has lots of associates with whom he regularly consults, and with whom he shares the glory.

Santosh, Mole's first impression is that the PDC is on the LAN, but the problem is that none of the servers can see it. Mole knows that this sounds somewhat contradictory, but there really is a difference. The second thing that Mole is curious about is whether the PDC and the BDCs are all in the same domain. Are there any trusts involved in this scenario?

Mole thinks that if you sit down at the PDC's keyboard and try to initiate a share connection to a remote machine (BDC or your favorite workstation) or connect to a printer, the good news is you would very likely be successful. The bad news is, there is a chance that you might fail. But, as in all things worth learning, at least the attempt would tell you something either way.

One little bit of trivia that was not included in your problem description (thanks for letting Mole know about the NT and Service Pack versions, by the way) was exactly How do you know that there is no PDC? Sometimes NT's error messages can be misinterpreted, causing IT Pro Support folks to leap to the wrong conclusion. Mole has seen this happen more than once.

One message that Mole is familiar with is the following:

A domain controller for your domain could not be contacted. You have been logged on using cached account information. Changes made to your profile since you last logged on may not be available.

This message is a result of the Netlogon service not being able to contact a Domain Controller before reaching its timeout setting. If, dear reader, this appears to be happening to you, then you can fiddle with the registry setting to increase the amount of time Netlogon waits before timing out during an interactive logon using a Domain User account. Please read Knowledge Base article: **163204:**Increase Domain Logon Timeout over Network for the details.

Another possible reason that the PDC is "not on the LAN" could be related to those pesky NATs. Not the small insect gnat that the Mole finds so tasty—rather the Network Address Translator kind of NAT.

Here is the situation: When you have a Network Address Translator (NAT) separating a Windows NT domain controller from its domain members or other trusted domains, Netlogon communication may fail. You will still be able to successfully redirect a drive across the NAT, and browse across the NAT, but logon attempts and trusts may fail. Why does this happen? Good question! Almost all the time it's because your NAT is not translating the source IP address from the NetBIOS header in your network traffic. Another KB article to read is 172227: Network Address Translators (NATs) Can Block Netlogon Traffic.

The other little bit of information that Mole picked out of your question was the phrase "sometimes there is no PDC on the LAN." That word "sometimes" is often the bane of network troubleshooters because those problems that occur in an apparent random fashion are the most difficult to track down and resolve. Mole has been there—and Mole empathizes.

So, to sum up, Mole recommends that you capture any error messages on the workstations that you receive, documenting the message exactly as it appears. Next, be sure to take a close look at the NT Event Viewer, System log for logged events at the time where your situation indicated that the PDC couldn't be found. Do this on both the workstation and the PDC. Also, note circumstances when the problem does not appear. Gather ammunition. Query the Knowledge Base using any NT Events that seem to be applicable. Prepare to get support from a troubleshooting professional. Good luck!

Mole

"Converting Raw Drives???"

Hello,

Is it possible to convert a raw drive?? I have been reinstalling NT in order to make the entire system NTFS. I would prefer to save time and just enter convert c: /fs:ntfs /v this would save me a lot of extra work. I am getting an error that the drive is in RAW format and cannot be converted. Hopefully it can some how.

Thanks for your help.

David Debreceni

Dear David,

Alas, despite his delight in all things raw, Mole must say that the answer is quite definitely "No." This is why.

First, let us review. A raw drive (or more properly a "raw partition") is one that has not been formatted with any file system (like FAT, FAT32, NTFS, etc.). UNIX folks are familiar with raw drives.

Now, Mole knows about applications such as SQL Server 7.0, which can write to a database file that resides on a raw partition. This is because SQL Server doesn't give a hoot about anything as un-database-like as a mere file system.

Now, the reason that you can't use NT's convert utility to convert the file system to NTFS is because the convert utility only converts a partition formatted as FAT to one formatted as NTFS. Only FAT to NTFS. Not FAT32 or HPFS or Raw to NTFS. Mole reiterates--just FAT to NTFS. This is a one-way street, however. Mole cautions you that—like going home again—once you've gone from FAT to NTFS, you can't go back from NTFS to FAT again.

One final cautionary note – there is a problem where the text mode part of setup stops responding (hangs) when installing Windows NT 4.0 on a system that has more than two RAW drives. Why is you ask? Mole is pleased you did, and hastens to answer. One requirement of Windows NT Setup is to determine that all of the drives have the same FT signature, same checksum, and same partition table status (TRUE or FALSE). The setup function "SppInitializeHardDiskArcNames" runs into a problem trying to match up the disks to the ARC names.

Mole can't really think of any good reason that you'd want to use raw drives (although you can mirror to a raw drive), the possible miniscule performance improvement, which is not a guarantee, just isn't worth giving up the ability to perform mundane file tasks such as copy, delete, or rename.

So, what should you do if you find yourself in the predicament of having more than two Raw drives? Simple: Reduce the number of Raw drives to two or less.

Hope this has been informative.

Regards,

Mole

Just how Many Backup Domain Controllers Are Enough?

Dear Mole,

I work for a medium size company (about 500 users, 10 production servers) that has five separate buildings that are networked through a fiber backbone. We are currently questioning the number of BDCs that should be on the network. There is one PDC and four BDCs (one in each building). One party feels that having that many BDCs is a waste of resources and that we don't need more than one BDC to backup the PDC. Users should be able to login from any building to the PDC without causing much chaos. The other side feels that segregating the buildings and making the users login at their specific server alleviates network traffic and bottlenecks at the PDC. Can you help us out with any suggestions?

Orlando

Orlando:

Mole is rarely shy about giving suggestions—but only when asked. You ask, so, here's Mole's thoughts:

I would say that the "first party"—the one that "feels that having that many BDCs is a waste of resources and that we don't need more than one BDC to backup the PDC" is correct. For 500 users distributed in four locations connected by fiber, one BDC is plenty. Mole also thinks that deciding where to place that BDC will depend on how those 4 locations are physically located regarding the LAN cabling. Probably placing it in the building where most users reside would be a good start.

Put those other BDCs to work as application, file, or print servers. Now, with only one BDC, you need to make absolutely sure that your disaster recovery procedures are rock solid - fully tested - and that the support staff understands them. Better yet, maybe you could use the extra servers for periodic disaster recovery testing and training for the support staff. Now there's an idea!

Mole

"SQL Server 7.0 Stored Procedures Give Bill a Headache"

Mole,

We use a SQL database to hold call details such as successful, unsuccessful and engaged calls and use a SQL Stored Procedure to return raw data to email to our customers.

However when we have added a considerable amount of data overnight the stored procedure hangs when we run the stored procedure.

I can get round the problem by running the stored procedure manually and stopping it after say 2-5 minutes, then run it again so it can recompile itself (it takes about 1 minute) and then run it one more time to make sure it has been optimized (it takes about 5-20 seconds).

I would ideally like to be able to run the stored procedure without my intervention. A similar stored procedure worked fine to SQL 6.5. Please can you help?

Bill Rushforth, Systems Development Manager

Hi Bill,

First be thankful that you're running SQL Server version 7.0. Not that 6.5 was a dog, mind you. Rather, 7.0 is a great product! Lots of really useful enhancements and performance improvements went into the 7.0 model. Uh-oh Mole is starting to sound like a marketer—and as Mole has reiterated over and over: he is not a marketer; he's a technical guy. In fact, Mole's whiskers are all a-quiver over the soon-to-be released SQL Server 2000. He can't wait!

Bill, Mole thinks he can help. Here are a few of ideas you might want to try:

  1. Run the Index Tuning Wizard against the table(s) that contains the most frequently updated data. Mole thinks that it wouldn't hurt to run the Wizard against all your tables (being a believer in the empirical method). And while you don't have to accept the Wizard's suggestions, Mole is of the opinion (along with the developers and a whole lot of SQL gurus) that this Wizard is worth its weight in gold—it is very good at what it does. At least by running it, you will have a comparison of which indexing scheme would be optimal. Compare SQL Server's Index Tuning Wizard's opinion with the indexing scheme that you've cooked up. Do some tests in the lab. Choose the best scheme. Query the SQL Server 7.0 Books Online using the query (what else) "Index Tuning Wizard" for the details.

  2. Execute the 'UPDATE STATISTICS' T-SQL command on the table that has had a relatively large amount of data has been added to a table(s).

  3. Execute the SP after recreating your stored procedure using the WITH RECOMPILE option (see the 'create procedure' T-SQL statement in SQL Server 7.0 Books Online).

  4. Use the WITH RECOMPILE option when you execute the SP—this is simply executed as follows

    SP_RECOMPILE stored_proc_name

  5. Execute the SP_RECOMPILE system stored procedure against your stored procedure. This forces a recompile of your stored procedure the next time it is run.

That's what Mole can come up with, with the information that he was provided. Try his suggestions and report back, please.

SQLey yours,

Mole

TCP/IP Subnetting in 25 Words or Less

Dear Mole,

My problem is: I'm studying for my MCSE, and have now come across the illustrious hurdle of passing TCP/IP!

I need "plain English" explanations of how subnetting works and how to answer those awkward questions. Whenever I get a question like what subnet mask will give you x hosts etc I can't seem to comprehend how to work it out!!

Your help is appreciated!!!

Clifford LevitanTechnical Support Centre

Clifford,

Mole congratulates you on taking on the TCP/IP Certified System Engineer test—knowing what it takes to pass this test is certainly not something to sneeze at. Mole thinks that there are possibly more acronyms within the TCP/IP world than anywhere else.

Mole also knows that to sit down and write an explanation of subnetting "in plain English" is an exercise in redundancy. Simply put, it's been done, done, and done again. Besides, the trick is to define exactly what "plain English" means. Could mean one thing to one IT Pro, and a whole 'nother thing to the next.

Be that as it may, your question intrigues Mole because it is an area that most IT Pros need to understand, and it helps to have a "plain English" explanation of how it all is supposed to work. Thus, Mole has done extensive digging and has come up with the following list of "more readable than most" TCP/IP information articles:

Mole is sure that other IT Pros have their own favorites, but this list—culled from Mole's sagging, overburdened bookshelf—includes Mole's favorites.

BackTalk

Hi Mole. Great page!

Does the Mole have any knowledge of ways and means that would allow increasing/avoiding/bypassing the 4 partition limit enforced by the NT disk administrator when using a RAID 5 configuration?

Mike Stapley, Senior DBA / Systems Analyst

Thanks for the compliment, Steve. And yes, Mole does have knowledge about why there's a 4 partition limit—but don't blame NT, please.

The information about primary partitions and an extended partition is contained in the Partition Table, a 64-byte data structure located in the same sector as the Master Boot Record (cylinder 0, head 0, sector 1). The Partition Table conforms to a standard layout that is independent of the operating system. Each Partition Table entry is 16 bytes long, making a maximum of four entries available. Each entry starts at a predetermined offset from the beginning of the sector,

Now, if this has only begun to pique your curiosity, Mole recommends you read the Windows 2000 Resource Kit or the Windows NT 4.0 Workstation Resource Kit for even more excruciating details.

Regards,

Mole