AddExtraHardDisk

From SME Server
Revision as of 03:44, 15 December 2008 by Christian (talk | contribs) (→‎Mount: adding a note on the implications of adding a mount with fsck in fstab.)
Jump to navigationJump to search


PythonIcon.png Skill level: Advanced
The instructions on this page may require deviations from standard procedures. A good understanding of linux and Koozali SME Server is recommended.


This HowTo are based on Michiel Blotwijk original.

Description

How to add an extra hard disk to an existing (running) SME server 7.x.


Warning.png Warning:
Adding a Single Disk (or non-raid disks): If you are adding only a single or non-raid disk then please note that the added harddrive will NOT take part of a RAID-array and hence get no redundancy.

If you do have a disk failure (or complete raid array failure) of this additional unit, then it can also halt the boot process. See the note on configuring fstab.

It's intended for special purposes only where you can benefit from mounting a dedicated harddrive to an ibay.



Warning.png Warning:
If you do a complete re-install at a later time, please recall that all available disks will be used at that time including these new additions. If you do a re-install and wish to preserve this data then ensure you remove these hard drives at that time.


Assumptions

  • It assumes adding one PATA (IDE) harddrive as master on the second IDE controller, but the process will be similar in other situations. More on that in step 2.1.
  • The harddrive will be named newdisk, this name will not be visible to the user.
  • You already have an ibay named IBAYNAME, this will be redirected to the new disk (together with its content). This name may be visible to the user (depending on the permissions).
  • You can replace the names newdisk and IBAYNAME with whatever valid name, just keep track of the names.

Notes about RAID

  • When you install SMEserver, ALL connected harddrives will be part of an RAID-array.
  • When you add a drive to a running system, it will not take part of anything unless you do something about it. This is the case in this HowTo. Note that it will NOT take part in any RAID array and hence, get no redundancy.
  • If you want the new drive to be part of the SME RAID array instead, get shell access, login as admin and select "5. Manage disk redundancy" and follow the instructions. The new disk should be empty and contain no partitions before doing this.

Step-by-step HowTo

Preparation

1.0 Mount the harddrive and fire up your machine.

1.1 Get direct shell access and login as root.

1.2 Tell init to go to single-user mode. (Optional, if the drives are complete clean, this is not necessary)

telinit 1

This will make you loose connection if you working through an ssh-client (like Putty). You need direct access to the shell to proceed.

Partition

Identifying the disks

2.0 Now you need to identify the device name of the new disk.

As a safety measure, identify and mark your disks. The following smartctl command will identify various parameters including the serial number which can be matched to the serial number on the physical package.

smartctl -i /dev/hdc 

The following command will show you all the drives on your server (including any USB drives you might have).

fdisk -l | more 
Partition overwiev
Role IDE controller Device name
Master 1 /dev/hda
Slave 1 /dev/hdb
Master 2 /dev/hdc
Slave 2 /dev/hdd


Important.png Note:
If you are setting up a raid array, it is a really good idea to eliminate single points of failure. For example try to avoid putting to disks of the same raid array on the same ribbon cable.


Paritioning the disk

2.1 Let's assume you installed the disk as master on the second IDE controller.
This means you have to partition /dev/hdc

fdisk /dev/hdc

and when prompted:

n (to add a new partition)
p (to make a primary partion)
1 (that's the number one, the number you want to assign to the partition)

Only if you will be adding this disk to to a RAID set, then you will also need to change the partition type.

t (to change partion type)
L (within t's prompt to list available parition type)
fd (normally the parition type is fd but verify from the L command and look for "Linux raid autodetect"

Accept the suggested first & last cylinder values

w (write and exit)

This will create the primary partition /dev/hdc1, using the entire disk space.

You will need to do this for each new disk you added.

Formating for a non-raid disk

2.2 For a non-raid disk, Format the new partition to ext3:

mkfs.ext3 /dev/hdc1


Additional steps to create a raid array from multiple disks

2.3 When you create an array, the various disks you added will be combined and be represented as a single entity typically identified as /dev/mdxx where xx is a number. For example, the SME Raids are typically on md1, md2, and in some cased md0.

For your purposes, you will choose an available one. In this example we will use md11 in an attempt to not get in the way of future changes to SME and make it easy to identify your new set.

First we need to add the newly (and identically) partitioned disks to the raid device:

mdadm --create --verbose /dev/md11 --level=1 --raid-devices=1 /dev/hdc1 /dev/hdd1

At this point the Raid will begin building.

Next you will need to format the new array:

mkfs.ext3 /dev/md11

You can verify that all is good with the following commands:

cat /proc/mdstat
mdadm --detail --scan --verbose

next you will need to edit the /etc/mdadm.conf file to add your array with the following after the other ARRAY entries:

# added by <your name here> -- so you can identify later you did this.
ARRAY /dev/md11

Alternatively you could add the output of

mdadm --detail --scan --verbose

to this file instead to be more specific. I have not tried this.

Mount

3.0 Create a mounting point for the new disk


Important.png Note:
This method uses fstab to mount your disk and forces an fsck. This method ensures that the disk is up and available early and has been verified for consistancy. However, the side effect is if there is a fatal error with the added single disk or the entire added raid array, then the system will stop at the fsck step in the boot process. If this happens, you will be given an option to log into the shell where you can perform recovery operations or modify fstab to prevent the disk from mounting and continue without the disk so that you can rectify the situation at your leisure while allowing the core system to run.


A mounting point is just an empty directory that will link to your new disk.
This can be any valid directory on your server, but it is recommended not to use /home or one of its subdirectories.
In step 4 you will see a much better method to link your /home directory to the second disk.

3.1 Create a directory underneath /mnt

mkdir /mnt/newdisk

3.2 Automount at boot time
To automatically mount the partition at boot time, you need to add the following line to the file /etc/fstab

This entry assumes among other things that you will be running user and group quota, the "1" indicates it is a disk that should be backed up with dump, the "2" indicates fsck should be run as a second priority. If you wish to modify these parameters, ensure you fully understand them as these are crucial to your system's operation.

It can be done in the text editor Pico

pico /etc/fstab

Add the following data to the file for a SINGLE disk (separate columns with a hit on the space-bar):

/dev/hdc1 /mnt/newdisk ext3 usrquota,grpquota 1 2 

or for a RAID array:

/dev/md11 /mnt/newdisk ext3 usrquota,grpquota 1 2

Make sure you end this line with a newline (with enter).
Hit <ctrl-x> to exit the editor, y and enter to save.

3.3 Mount manually (this time) all filesystems listed in /etc/fstab

mount -a

3.4 Report the amount of free disk space available on all mounted filesystems + the type of each filesystem, check if it looks OK.

df -T
Filesystem Type 1k-blocks Used     Available Use% Mounted on
/dev/hda3  ext3 39151724  28441348 8721532   77%  /
/dev/hda1  ext3 101089    8953     86917     10%  /boot
/dev/hdc1  ext3 38474956  27948    38447008   0%  /mnt/newdisk

If you don't see a line for /dev/hdc1 (or /dev/md11 if it is raid), something went wrong.
Go back to step 2 and check if you followed all instructions correctly.

Mapping

4.0 Alternatives

Alternative 1: Use /home to mount the new disk

On standard Linux machines, the /home directory contains the user's directories and is therefore a popular place to mount a bigger disk.

However, on a SME server this is not a good idea:

SME stores all its configuration files in /home/e-smith/db.
If for whatever reason your disk doesn't mount, the configuration files will no longer be accessible, causing all kinds of nasty problems.

Mounting your second disk to /home or even /home/e-smith/files effectively means that your first disk will no longer be used to store user files.
In some cases that might be a good idea, but in most cases you are just wasting the remaining space of your first disk.

Alternative 2: A much better approach is to leave the second disk mounted at /mnt/newdisk and to create symlinks (symbolic links) for specific ibays or user folders.

The following example shows you how to place the IBAYNAME ibay on the second disk.

4.1 Create the necessary subdirectories on your second disk

mkdir -p /mnt/newdisk/ibays/IBAYNAME


4.2 Check the ownerships and permissions on the original directory

ls -l /home/e-smith/files/ibays/IBAYNAME

drwxrwxr-x 18 admin allstaff 4096 Sep 23 14:48 IBAYNAME

4.3 Change ownership as in the original. The "-R" (optional) means recursive, subdirectorys will be included in the command.

chown -R admin:allstaff /mnt/newdisk/ibays/IBAYNAME

4.4 Set permissions to the ibay IBAYNAME as in the original. For help on this, look at bottom of this page. The "-R" (optional) means recursive, subdirectorys will be included in the command.

chmod 775 -R /mnt/newdisk/ibays/IBAYNAME 

4.5 Copy the content to your new disk

cp -rip /home/e-smith/files/ibays/IBAYNAME/* /mnt/newdisk/ibays/IBAYNAME 

4.6 Check if all files have been copied correctly

diff -r /mnt/newdisk/ibays/IBAYNAME /home/e-smith/files/ibays/IBAYNAME 

4.7 Remove the original files

rm -rf /home/e-smith/files/ibays/IBAYNAME 

4.8 Create a symlink to the new location

ln -s /mnt/newdisk/ibays/IBAYNAME /home/e-smith/files/ibays/IBAYNAME 

4.9 Test the new location

ls /home/e-smith/files/ibays/IBAYNAME

4.10 If you want the ibay to be accessible from the internet, you also need to enable "follow symlinks" in apache:

db accounts setprop IBAYNAME FollowSymLinks enabled
signal-event ibay-modify IBAYNAME

Quota (optional)

5.0 Next you might want to activate the quota on the disk.
This will only work if you activated quota management for at least one user (Server Manager > Collaboration > Quota Management).

quotacheck -vugc /dev/hdc1

If you get "quotacheck: Old file not found" you probably haven't activated quota management.
If quotachecks comes back to the command prompt without errors, you're ready.

Final check

Reboot your machine

reboot

Check if the drive was mounted correctly with

df -T

and/or surf to your ibay.

If it looks fine, relax and tell the world about it, you made it!

Notes about file permissions

  • For help about chmod and file permissions, look at

ss64 or oreillynet.