Server Fault is a question and answer site for system and network administrators. It only takes a minute to sign up.
Sign up to join this community
Our previous sysadmin created several partitions inside one hard drive in the fileserver, saying that it minimizes corruption:
In case one partition goes bad, it doesn't affect others.
Question: I've never seen a partition goes corrupted, so is it true?
Machine: Dell R200, 1TB hard drive with 6 partitions (that he created), Windows Server 2003
Please note that this server is just used to serve files & also store ex-staff files.
For the most part a 'corrupt partition' is actually referring to a corrupted filesystem. Filesystems are contained within partitions(1). So if the filesystem in one partition gets corrupted, the filesystems in different partitions on the same disk will not be. That said, the same event (abrupt power-down, others) that corrupted one filesystem can also corrupt other filesystems on the same disk.
Actually corrupting a filesystem takes some doing these days. I've seen some pretty bad NTFS pukes, but they were all related to exceptional events. A good way to cause problems is to yank the powercables out the back of a Windows server while it's doing heavy I/O. For SAN-attached storage (not your problem) sometimes if the SAN hiccups for some reason the OS can react strangely; I had a case on Server 2003 where just such a hiccup caused Windows to flag about 20GB of the partition as bad-clusters (they weren't, a chkdsk /F /R removed the bad-flag from those clusters). Abrupt power loss is probably the most common cause of such problems. Having your drive partitioned so that your heavy I/O applications are on a different partition than your OS boot environment means that when the power is yanked the chances of boot being affected are much smaller.
(1) Unless they're part of a volume management system like LVM, or Microsoft Dynamic Disks, at which point things get a bit fuzzier.
Yes. It sounds like the previous sysadmin was probably concerned with corrupt filesystems (which are built on partitions), rather than the partition table itself. And filesystems definitely can be corrupted, and splitting them is a traditional measure to reduce the risk.
Additionally, smaller filesystems are faster to fsck (or chkdisk or whatever the Windows equivalent is), so when something goes wrong, downtime is reduced.
Six arbitrary partitions may cause more administrative overhead than it is worth, however, particularly if there's no good way to resize them on the fly.
Ywes, I have seen that repatedly. THe main problem withh standard partition tabelsi s that there is ONE - this is different for example for more modern partitioning technologoies. If that one sector in the disc dies, the partition nifo is just toast.
