My root filesystem is
Submitted by Anonymous on Thu 10 Aug 2006
| Network mounted |
![]() 2% | 14 votes |
| Software RAID |
![]() 14% | 97 votes |
| Hardware RAID |
![]() 9% | 61 votes |
| LVM |
![]() 10% | 71 votes |
| Plain/Normal |
![]() 64% | 432 votes |
| Total 675 votes |
[ Parent ]
[ Parent ]
The only problem I see with LVM is : it's easy to do a mistake and loose a lot of data :/ it really need some practice on a non critical system for a few month imho.
[ Parent ]
I therefore have my root mounted over NFS.
The machine runs LVM over software RAID.
This is so that I can use a fanless laptop as my client and the huge grunting, sweating server lives in the former coal hole and my wife doesn't complain about the noise.
Surely this is not so uncommon.
[ Parent ]
My desktop system is ext3 on plain PC partitions on a single SATA disk.
My test system has all filesystems bar /boot in a LVM2 volumegroup which spans two old IBM disks. It's not using LVM2 mirroring or RAID. Other than an initrd issue some time ago, now fixed, it's been perfectly okay.
My home server is currently ext3 on plain PC partitions, on several small old EIDE disk but I plan to move most of the file systems onto xfs on LVM2 this weekend.
Eventually I plan to build/buy a new home server, and use RAID/LVM2 and export it's file system to all my other systems.
--
"It's Not Magic, It's Work"
Adam
[ Parent ]
I've been using XFS for about a year on the setup I describe above (the coal hole server), but it died with lots of I/O errors and "Unknown error (990)" on Sunday. Couldn't get the box to stay up for longer than about five minutes.
The only thing that I could find was dated April 2004 (http://oss.sgi.com/archives/xfs/2004-04/msg00130.html).
This points to this SGI bug:
http://oss.sgi.com/bugzilla/show_bug.cgi?id=323
which is still marked as NEW.
Tried xfs_repair and it gave the same list of problems each time, which suggests to me that it was not able to repair and I couldn't save the log reliably (for some reason the beginning or end of the file would be missing and xfs_repair would give the statement "Stopped" rather than going on to the end.
This disk contains my root, boot, tmp, usr, and var paths, so I rebuilt the box with Ext 3 on Monday evening and have had no problems since.
Although this is too short a time to definitely say that it is not a hardware issue, I am a little wary.
[ Parent ]
My Sarge box is still using 2.6.8, so I don't need to worry about 2.6.17 per se yet.
--
"It's Not Magic, It's Work"
Adam
[ Parent ]
--
"It's Not Magic, It's Work"
Adam
[ Parent ]
Everything else I have under LVM.
/boot and / (root) are ext3 - using reiserfs for all other
LINUX filesystems (at least my Debian systems).
Not much to mirror to on single hard drive sytems (e.g. laptop).
Someday I'll get even snazzier, ... but that day's not today.
... etch --> stable ... LUKS? :-) ...
also planning some laptop upgrades in the not-too-horribly-distant-future.
Eventually I'll also have a hot swappable 2nd hard drive, so there may well
be some RAID-1 intermittently active there (and/or some fairly frequent backups
of more critical stuff).
I also plan to eventually get my desktop system from one hard drive, to
three - and to include various software RAID (probably mostly -5 and some -1,
and perhaps a wee bit of -0)
Hardware RAID *can* be great (very nice on some hardware) ... it can
also be sucky (there are some very poor hardware RAID implentations out
there too). But for small (2 or 3) disk systems, it's not nearly as
flexible as software RAID (e.g. with 2 or 3 disks, most hardware RAID doesn't
allow mixes of RAID-1, RAID-5, and RAID-0 and/or non-RAID to all be present
simultaneously on those drives. For some systems/applications, that might
not matter, but for others it's important/critical.
http://www.rawbw.com/~mp/linux/lvm/balug/
[ Parent ]
[ Parent ]
At work most of the servers are software RAID 1 pairs, with ReiserFS 3, and a couple of servers with email or database roles have hardware RAID (although in general on the kind of cheap hardware it is on, I now think it would have been wiser to use software mirroring, enable the write cache on the disks, and take the data loss as and when the UPS gives up) and ReiserFS 3.
The XFS stuff moans elsewhere are either a smelly recent patch, or bad luck. Most of these filesystems do exactly what they say on the tin. Me I'd only deploy XFS where I needed its online snapshot capability, horses for courses.
[ Parent ]

2%