RedHat Linux and emcpower files

[J117]

RedHat Enterprise 5.4 (Tikanga)
Fibre disks presented by EMC Clarion.

In my environment one Red Hat server has its Clarion devices mounted in two different ways within fstab.

/dev/emcpowera1 /usr1 ext3 defaults 0 0
/dev/emcpowerc /usr2 ext3 defaults 0 0
/dev/emcpowerb /usr3 ext3 defaults 0 0

All three file systems are usable and seem to be performing well.

Is it incorrect to create file systems directly at the device level without using a partition? (/dev/emcpowerc)??

Are both methods acceptable?
Is there any downside to not creating the partition?

I will never need to carve these luns up. The entire disk is dedicated to a single file system. In two of these cases the file systems are a TB or larger.

I am particularly concerned if there is any danger to the data or a performance hit of any sort.

Thank you.

[Steven E. Protter]

Shalom,

You are better off using lvm after emc is properly installed.

Using lvm enables you to add a lun to the volume group and extend the file system without downtime.

partition with fdisk
pvcreate
vgcreate
lvcreate
mkfs.ext3

SEP

[Matti_Kurkela · Accepted solution]

Both methods are workable. However, I would recommend using LVM instead.

When you don't have a partition:

+ you are not restricted by the 2 TB hard maximum size limit of the traditional PC partition table

- if the SAN administrator makes a mistake and accidentally presents the same disk to a Windows host too, the disk will look like unpartitioned space to Windows

- you cannot migrate your data to another disk (or disks) on-line without stopping the applications and unmounting the filesystem

- maximum filesystem size can be no greater than the size of a single disk/LUN

? expanding the filesystem on-line may be possible, if you increase the size of the LUN and use the "partprobe" command to make the system re-detect the LUN size. However, changing the size of active LUNs is risky.


When you do use a partition:

+ if the SAN administrator accidentally presents the disk to a Windows host, Windows can detect the disk is already partitioned and obviously in use

- expanding the filesystem is even more difficult than in the previous case, with the same risks

- you cannot migrate your data to another disk (or disks) on-line without stopping the applications and unmounting the filesystem

- with the traditional PC partitioning technology, 2 TB is the absolute maximum disk/LUN size

- maximum filesystem size can be no greater than the size of a single disk/LUN


When you use LVM:

+ you are not restricted by the 2 TB hard maximum size limit of the traditional PC partition table

? if the SAN administrator makes a mistake and accidentally presents the same disk to a Windows host too, the disk will look like unpartitioned space to Windows
(but if you worry about this, you can use a partition table with LVM too)

+ on-line data migration is possible: you can move your data from one storage system to another while filesystems are mounted and applications are using them

+ you can make a volume group out of multiple physical volumes, and create a single large logical volume over all of them = you can have filesystems larger than any single disk/LUN

+ on-line extending filesystems is easy: add another disk/LUN to the same volume group as the disk you need extended, then expand your logical volume to use it too, then expand the filesystem to use the new space


In my opinion, the ability to migrate the data from one storage to another while it's being used is very important: after 3-5 years, your Clarion will change from a "shiny new thing" to "old clunker for which is impossible to get a reasonably-priced support agreement any more". At that point, you will have a new, bigger storage system and you'll want to move your data to it and get rid of the old one. That will be *much* easier if you can migrate the data with no downtime.

MK