Frankly speaking, you cannot create a Linux partition larger than 2 TB using the fdisk command. The fdisk won't create partitions larger than 2 TB. This is fine for desktop and laptop users, but on server you need a large partition. For example, you cannot create 3TB or 4TB partition size (RAID based) using the fdisk command. It will not allow you to create a partition that is greater than 2TB. In this tutorial, you will learn more about creating Linux filesystems greater than 2 Terabytes to support enterprise grade operation under any Linux distribution.
To solve this problem use GNU parted command with GPT. It supports Intel EFI/GPT partition tables. Partition Table (GPT) is a standard for the layout of the partition table on a physical hard disk. It is a part of the Extensible Firmware Interface (EFI) standard proposed by Intel as a replacement for the outdated PC BIOS, one of the few remaining relics of the original IBM PC. EFI uses GPT where BIOS uses a Master Boot Record (MBR).
(Fig.01: Diagram illustrating the layout of the GUID Partition Table scheme. Each logical block (LBA) is 512 bytes in size. LBA addresses that are negative indicate position from the end of the volume, with −1 being the last addressable block. Imaged Credit Wikipedia)
Linux GPT Kernel Support
EFI GUID Partition support works on both 32bit and 64bit platforms. You must include GPT support in kernel in order to use GPT. If you don't include GPT support in Linux kernelt, after rebooting the server, the file system will no longer be mountable or the GPT table will get corrupted. By default Redhat Enterprise Linux / CentOS comes with GPT kernel support. However, if you are using Debian or Ubuntu Linux, you need to recompile the kernel. Set CONFIG_EFI_PARTITION to y to compile this feature.
File Systems
Partition Types
[*] Advanced partition selection
[*] EFI GUID Partition support (NEW)
....Find Out Current Disk Size
Type the following command:
# fdisk -l /dev/sdb
Sample outputs:
Disk /dev/sdb: 3000.6 GB, 3000592982016 bytes
255 heads, 63 sectors/track, 364801 cylinders
Units = cylinders of 16065 * 512 = 8225280 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x00000000
Disk /dev/sdb doesn't contain a valid partition tableLinux Create 3TB partition size
To create a partition start GNU parted as follows:
# parted /dev/sdb
Output:
GNU Parted 2.3 Using /dev/sdb Welcome to GNU Parted! Type 'help' to view a list of commands. (parted)
Creates a new GPT disklabel i.e. partition table:
(parted) mklabel gptSample outputs:
Warning: The existing disk label on /dev/sdb will be destroyed and all data on this disk will be lost. Do you want to continue?
Yes/No? yes
(parted)Next, set the default unit to TB, enter:
(parted) unit TBTo create a 3TB partition size, enter:
(parted) mkpart primary 0 0OR
(parted) mkpart primary 0.00TB 3.00TBTo print the current partitions, enter:
(parted) printSample outputs:
Model: ATA ST33000651AS (scsi) Disk /dev/sdb: 3.00TB Sector size (logical/physical): 512B/512B Partition Table: gpt Number Start End Size File system Name Flags 1 0.00TB 3.00TB 3.00TB ext4 primary
Quit and save the changes, enter:
(parted) quitSample outputs:
Information: You may need to update /etc/fstab.
Use the mkfs.ext3 or mkfs.ext4 command to format the file system, enter:
# mkfs.ext3 /dev/sdb1
OR
# mkfs.ext4 /dev/sdb1
Sample outputs:
mkfs.ext4 /dev/sdb1 mke2fs 1.41.12 (17-May-2010) Filesystem label= OS type: Linux Block size=4096 (log=2) Fragment size=4096 (log=2) Stride=0 blocks, Stripe width=0 blocks 183148544 inodes, 732566272 blocks 36628313 blocks (5.00%) reserved for the super user First data block=0 Maximum filesystem blocks=4294967296 22357 block groups 32768 blocks per group, 32768 fragments per group 8192 inodes per group Superblock backups stored on blocks: 32768, 98304, 163840, 229376, 294912, 819200, 884736, 1605632, 2654208, 4096000, 7962624, 11239424, 20480000, 23887872, 71663616, 78675968, 102400000, 214990848, 512000000, 550731776, 644972544 Writing inode tables: done Creating journal (32768 blocks): done Writing superblocks and filesystem accounting information: done This filesystem will be automatically checked every 31 mounts or 180 days, whichever comes first. Use tune2fs -c or -i to override.
Type the following commands to mount /dev/sdb1, enter:
# mkdir /data
# mount /dev/sdb1 /data
# df -H
Sample outputs:
Filesystem Size Used Avail Use% Mounted on /dev/sdc1 16G 819M 14G 6% / tmpfs 1.6G 0 1.6G 0% /lib/init/rw udev 1.6G 123k 1.6G 1% /dev tmpfs 1.6G 0 1.6G 0% /dev/shm /dev/sdb1 3.0T 211M 2.9T 1% /data
Make sure you replace /dev/sdb1 with actual RAID or Disk name or Block Ethernet device such as /dev/etherd/e0.0. Do not forget to update /etc/fstab, if necessary. Also note that booting from a GPT volume requires support in your BIOS / firmware. This is not supported on non-EFI platforms. I suggest you boot server from another disk such as IDE / SATA / SSD disk and store data on /data.
Further readings:
- How Basic Disks and Volumes Work (little outdated but good to understand basic concept)
- GUID Partition Table from the Wikipedia
- man pages parted
Updated for accuracy!
Featured Articles:
- 20 Linux System Monitoring Tools Every SysAdmin Should Know
- 20 Linux Server Hardening Security Tips
- Linux: 20 Iptables Examples For New SysAdmins
- My 10 UNIX Command Line Mistakes
- 25 PHP Security Best Practices For Sys Admins
- The Novice Guide To Buying A Linux Laptop
- Top 5 Email Client For Linux, Mac OS X, and Windows Users
- Top 20 OpenSSH Server Best Security Practices
- Top 10 Open Source Web-Based Project Management Software
Facebook it - Tweet it - Print it -
We're here to help you make the most of sysadmin work. So, subscribe!
{ 45 comments… read them below or add one }
Typically with Linux and large disks, I prefer just to use LVM directly on the disk, no partitioning at all. I find this to be very stable and manageable.
Ryan, how do you tell grub to load the kernel from a lvm partition? IMHO, grub wont mount logical volumes and thus, cannot find your kernel.
Thanks for the CLI verify – I used gPartED for the same – with graphical confirmation;)
Using the default distro kernel in Debian 4.0 (etch):
$ uname -a
Another tip, instead of doing
# mkpart primary 0 3001G
Do
# mkpart primary 0 -0
This will tell parted to fill up the entire drive with one volume.
I could create on 3TB volume following guide above. But when rebooting, the pratition tabled is wrong. Pleas Help /T
when i’m booting my own build debian kernel,
it stocks when the system want’s to mount to root filesystem.
it says somthing like
the following disks found
0800 sd sda
0810 sd sdb
the error is
VFS: cannot open root device “sda2″ or unknown-block(8,2)
Please append a correct “root=” boot option; here are the available partitions:
0800 244129792 sda driver: sd
0810 9765519360 sdb driver: sd
All supported releases of Ubuntu (ie 6.06 LTS and higher) do have ‘CONFIG_EFI_PARTITION=’y compiled into the kernel.
Since recent versions of Ubuntu and Debian have CONFIG_EFI_PARTITION compiled into their stock kernels, you should remove the part about having to recompile the kernel so that readers don’t needlessly recompile their kernels.
What tripped me up is mkpart gpt is actually a lone command followed by a sub-command.
Don’t type them both at once, wait for the next prompt before typing gpt.
With EXT3 on most systems the max block size is 4096 bytes, which will still limit you to ~8TB for your filesystem.
Hi,
How to Create a Partition Size Larger than 2TB in Freebsd 7.2 64-bit and 32-bit OS. I have a hardware raid 5 disk contains total 8TB of size and want to create a partition like one 5TB and one 3TB.
How I can get it done in FreeBsd.
Thank you
Manish
i don’t understand the premise.
since i can’t buy a > 2tb drive, i don’t see the problem.
fdisk /dev/sda
(make a partition of type fd, that spans the entire 2tb drive)
fdisk /dev/sdb
ditto sdc
ditto sdd
mdadm (assemble sda through sdd into a raid 5 of size 6 terabytes)
then mkfs /dev/md0
show me where with the biggest drives you can buy as of today, that you’ll have to run “fdisk” against something larger then 2tb????
Clearly you never worked in an enterprise setup where you can put 24 x 1TB size drive put in RAID mode.
Thanks for the great tut. There is 1 basic part I could not understand, once you create a single 3 TB partition using “mkpart primary 0 3001G”, how do you create partitions for OS, swap and other stuff.
Great Post!!!
Prior to this article, I was having trouble partitioning a 6TB Raid 5 Array (4 x 2TB disks).
I’m running CentOS (2.6.18-164.10.1.el5) and used the following commands from this article and comments. Thanks!
————————-
parted /dev/sda
mklabel gpt
mkpart primary 0 -0
quit
mkfs.ext3 /dev/sda
* Note: I first tried using “gparted” but something with the data types in the graphical interface didn’t like the size of the raid array and was displaying negative values. Stupid gui.
Just for people who get search hits on this page…I suggest the commands:
parted /dev/sda
mklabel gpt
mkpart primary 0 -0
print
quit
mkfs.ext3 /dev/sda1
SassyZ needs to understand that file systems are created within partitions while GPT are applied to disks. In other words, “parted mklabel gpt /dev/sda” writes a partition table to the first few sectors of a disk drive. The “mkpart” command creates a partition entry in that GPT. The first partition is named “/dev/sda1″. So the mistake in the last command is specifying the wrong device special file name. The command should be “mkfs.ext3 /dev/sda1″ to create a file system in the first partition. The command “mkfs.ext3″ will create a file system starting at sector zero and overwrite the GPT. (Note I’ve never tried a size of -0, so use print to verify the partition that was created.)
[quote]show me where with the biggest drives you can buy as of today, that you’ll have to run “fdisk” against something larger then 2tb????[/quote]
Try a hardware RAID for a change. The RAID controller will report a single huge drive to your OS. I just skip partitioning the RAID device altogether. *shrug*. Just mkfs /dev/sdb.
No fiddling with exotic partitions. Not the prettiest of solutions, but very reliable nontheless.
This isnt working for me.
(parted) mklabel
New disk label type? gpt
Error: Invalid argument during write on /dev/sdb
Retry/Ignore/Cancel?
Would this have anything to do with it…
Warning: Device /dev/sdb has a logical sector size of 4096. Not all parts of
GNU Parted support this at the moment, and the working code is HIGHLY
EXPERIMENTAL.
If so, how can I get around this? I’m using four 2TB drives in RAID5 – volume size is 5.9TB – it is ISCSI.
[QUOTE]I just skip partitioning the RAID device altogether. *shrug*. Just mkfs /dev/sdb[/QUOTE]
I tried that but if I run fdisk -l i get….
Disk /dev/sdb doesn’t contain a valid partition table.
Will this cause any problems?
No. Not unless you try to read the drive with a different operating system or reinstall the operating system. Without the partition table, other OSes will think the drive is unpartitioned (blank). As long as you keep using Linux or perhaps BSD, just tell the system that the drive is a big single partition (mount /dev/sdb -t ext3 /mnt/bigdrive -o noatime) and it should work fine.
If you want to protect the drive from another OS (I don’t know why you’d dual boot with a massive RAID like that) or make it more portable then you should follow the instructions in the article for creating a GPT instead.
Excelent ! works as a charm
The BIOS does NOT use the MBR. It simply loads the first sector of the disk into memory at 0x7c00, and runs whatever is there. IBM way back when decided to use an MBR to partition drives. The bios knows nothing about the partitions or how the MBR is layed out, and can perform perfectly without an MBR. So much mis information for such a small article. You can use a GPT with a standard bios, it’s just that EFI is ‘aware’ of the GPT, where a standard bios knows nothing of it, and just loads the first sector of the disk, and lets it do it’s thing. All they did was move some code from the hard drive into the firmware, doesn’t really make any difference where it resides, it’s all software that performs the same function
I thought I setup everything correctly but I am also getting
Disk /dev/sdb doesn’t contain a valid partition table.
The drive seems to work correctly but I dont like to see errors even if it is a false positive
Joe
ATL support systems
678-243-5712
A common message as there is not partition table. This also displayed when you have software raid devices such as /dev/md0
Nothing to worry. Just ignore warning.
i have an ISCSI 4TB storage, i have partitioned it as needed with GPT label, but suddenly after almost a year of usage when it reached now about 2.5TB i get those messages in the kernel:
kernel: attempt to access beyond end of device
its debian lenny, parted looks good, no errors.
Model: IBM 1726-3xx FAStT (scsi)
Disk /dev/sdb: 3999GB
Sector size (logical/physical): 512B/512B
Partition Table: gpt
Number Start End Size File system Name Flags
1 17.4kB 3999GB 3999GB ext3 primary
any suggestions?
Awesome! Thanks for this, after tweaking fdisk settings for ages i came accross your article. Sorted :D
I used parted as described to work with large arrays, so I am quite happy with it. One question though. What would be the easiest way to copy partition tables between disks (not partitions!). I know in fdisk there was a tool called sfdisk that let you dump / rewrite partition tables from one disk to another…. is there a similar procedure in parted? any hints greatly appreciated! Adrian
GPT stands for GUID Partition Table. This mean that each partition is assigned a unique ID when the partition is created. So, copying the partition table is a BAD IDEA, even if you know to also copy the backup partition table at the end of the disk. You should run parted (or whatever) on each disk that needs the “same” partition table and create a new partition table with the same partition types and sizes, but with new (unique) GUID values.
Although not well documented, parted will accept starting location and size in sectors, if values in MB are being rounded and causing slightly different values than desired. To repartition the fourth disk sdd (and lose all of its content):
parted -s /dev/sdd mkpart fat16 34s 1048609s
Note that I’m worried about duplicating the GUID that uniquely identifies each partition, not the standardized GUID value that identifies the partition type.
Great Tut as always.
One question though… All this assumes you’ve got a running system to which you plan to add a second >2TB device. When using GPT, most modern linux distro’s will do as you described.
However, how do you install to such a device?
Situation = 8x2TB raid6 (hardware), and boot-able debian 6.0 cd -> The installer recognizes a 500 GB disk (I previously GPT’ed the device & partitioned it to my liking with gparted livecd).
Suggestions?
TIA,
Jord.
Solution for my original problem: ubuntu install cd (in my case 10.10) will only create GPT when the drive is large enough to require it, otherwise it will stick to msdos mbr.
So: make you’re raid as you wish it to be and install straight from cd.
One smaller problem I’m facing now is also GPT-related:
I installed ubuntu on 3x2TB hw raid whilst there still was a 5X2TB soft raid in the same machine. Next I moved the data from the soft raid to the hw raid and expanded the hw raid to its full 8x2TB in raid 6.
When now I want to enlarge /dev/sda3 (ext4) to full raid size I get blocked of by the GPT backup table <>
But I don’t quite see how to do that.
Regards,
Jord
quote missing in previous reply:
The backup GPT table is not at the end of the disk, as it should be. This might mean that another operating system believes the disk is smaller. Fix, by moving the backup to the end (and removing the old backup)?
Useful…
Thanks a lot…
If you get this…
Warning: The resulting partition is not properly aligned for best performance.
You really should run parted like so…
parted -a optimal /dev/sdb
I running problem on this. I try to mount a 4*2T RAID5 array with USB connect to NAS IOMEGA IX2-200, the IX2-200 is running Debian Lenny. How could I know if GPT is enabled?
I applied mklabel gpt successfully, but I still see my RAID5 as 2T disk.
how to create a mount point of 20 TB LUN in RHEL 5.4 32 bit?
When you get ready to format your new huge partition, I recommend using “-m0″ with the mkfs command. This will give you the maximum amount of space to use.
The ‘-m’ sets the reserved block to zero (0).
If you do not add that to the command, you will loose (i think) either 5 or 10% of the drive to reserved blocks.
# mkfs.ext3 -m0 /dev/sda1
@All,
I’ve updated the tutorial with step by step instructions.
I’m getting the same errors from Parted as mentioned by Chris above…
“Warning: Device /dev/sdd has a logical sector size of 4096.” and “Error: Invalid argument during write on /dev/sdd Retry/Ignore/Cancel?” At which point only Cancel does anything.
In my case the drive is a Hitachi XL3000 3TB (external USB 2.0) and I wanted to format it ext3. Since Parted didn’t work for me, I partitioned it under OS X (Disk Utility) as a single 3TB partition. And then mkfs.ext3 and tune2fs -L for the label. Seems to be working now, but if someone smarter than me wanted to explain how to do it without resorting to OS X, that would be awesome.
Any help?
Sean,
I seem to have just purchased the same drive you are talking about, and yes I had the same problem. I ended it up there by making an error in following the instructions above. If you look at the drive, you will notice two partitions on it. Both zero in size. Use rm command for each partition and go carefully back to the beginning.
Note, though, that the above instructions result with some empty space at the beginning of the drive. If you select units as Bytes, the empty space will be moved to the end of the drive.
Hi Nick,
Thanks for the reply. While in Parted, I still can’t get anywhere because the disk’s label remains unrecognized, and the mklabel command simply returns the above-mentioned error (pretty much every command in Parted returns the same error: “unrecognised disk label”). Should I be using a different Linux tool to look at the drive?
>Gadgetman October 16, 2008
>I could create on 3TB volume following guide above. But when rebooting, the pratition tabled is wrong. Pleas Help /T
I’ve just found the cause of this problem.
http://www.linuxquestions.org/questions/showthread.php?p=4519659#post4519659
Maybe reply #9 explains the reason.
Could anyone tell me whether this issue has already resolved or not.
DMC
Now the other question, how long will it take to fsck one of these 3TB filesystem? This could take many hours. We’ve seen it take over 12 hours for a large filesystem.
By default the OS will automatically fsck the filesystem “every 25 mounts or 180 days, whichever comes first”. This could be a serious issue if you want to reboot the system 180 days from now, as your 4 minute reboot could turn into a 4+ hour wait for fsck to complete.
If you want to disable the automatic fsck upon boot, try this command:
`tune2fs -i 0 /dev/sdb1` will also disable the automatic fsck.
-= Stefan
Every time i need some tech advise i go to google to find information, every other site is kind of scrambled, no clear information, until i find this page.
NOTHING needs to be explained…..everything clear as water.
Thank you so much guys, keep up the GREAT work