best practice for logical drive partition on a single disk

Tue, 10 Jun 2008 02:12:10 Z

We have a number of W2K3 Standard Edition servers. Each server has one disk and typically partitioned into c: (system), e: (app files) and f: (app data) drives. The main argument supporting partition this way instead of a single partition (c:) is that during an attack such as DOS when log file can be filled up quickly to cause disk full, it's possible to leave c: drive not filled up by putting log files to a non-system partition, say f: drive. So the system won't crash. However, since we are running out of space on certain drives due to sheer data volume (and suppose adding more disks and converting the drive to a spanned volume is out of the question) we have to re-evaluate this practice. Apparently single partition can maximize usage of physical disk. Besides, when a non-system partition filled up the services provided by the server is down, and a running OS that cannot provide the services is good for nothing.
The key question is therefore what happens when system drive c: runs out of space and whether it take more time to clean up system volume than non-system volume? If the system will crash (blue screen etc) and won't reboot successfully when c: drive is full, then multiple partition still has its merits; otherwise the argument is probably not justified.
I cannot seem to find documentation on neither the best practice for logical drive partition nor behaviors when system partition is out of space.
Any advise?
Thanks

best practice for logical drive partition on a single disk

Tue, 10 Jun 2008 07:25:23 Z

I don't know about any best practice, but talking to my colleagues we all agreed on this:

C: should only contain the OS, NOTHING else. Around 40-60 GB for this. If you need to retain the Windows security/system/application log you should send those to a central location, otherwise 5 mb worth of logs before overwrite seems to be standard. There are several third party products for syslog usage in Windows.

Logfiles from other apps and applications should be stored on a separate partition / multiple partitions.

If your C: is full and all you have on it is the OS it'll take a lot longer "cleaning" that up compared to any other drive since you just can't randomly delete files on C: for releasing space. For space maximization I'd recommend shared storage, SAN or NAS, which'll let you have apps and data centrally stored and your servers just configured with OS.

Technet Forums Moderator | Solution Specialist | Ask The Experts IT-forum

best practice for logical drive partition on a single disk

Tue, 08 Jul 2008 19:18:41 Z

Although it may sound a reasonable rule that C: should only contain OS and nothing else, in practise the rule can hardly be enforced. For example, user profiles are stored in c:, many software can only be installed in c: and even if they can be tweaked to install elsewhere they are vulnerable to error when deviated from the default installation config. Even OS itself is a growing organism of which the size depends on what optional features to install and cannot be easily foreseeable at the time disk is partitioned and core OS is being installed.
Worst of all, it is hard to extend system drive because means to expand a logical volume such as converting a basic disk to dynamic disk for stripped or spanned volume cannot be applied to system volume.
Therefore maximizing disk utilization means maximizing c: drive size, which translates to avoiding partitioning a physical drive that contains c: volume to more than one volume.
The answer to the question "what happens when system drive c: runs out of space" is nothing drastic. By drastic I mean things like blue screen of death, frozen screen, failed to reboot etc. The answer implies that the time it takes to recover from a filled-up server that contains only the c: drive is no more than that of a system with filled-up non-system drive. By recover I mean deleting those over-flooded files (log file for example) that caused disk full rather than "randomly delete files on C:".
This implication renders the main argument that favors multiple partition, i.e. easier to recover from a filled up disk, unjustified.

best practice for logical drive partition on a single disk

Tue, 22 Jul 2008 10:29:40 Z

The number of user profiles on a server is surely limited both in numbers and size. The number of users able to logon should be limited and profiles on a server shouldn't contain any big files or application settings anyway.

As far as application installations go I haven't seen an application in a few years time that doesn't allow you to change the installation path.

Technet Forums Moderator | Solution Specialist | Ask The Experts IT-forum