Difference between revisions of "Set up a two-server GlusterFS array Tutorial"

(Created page with "For Ubuntu 16.04. For Hard drive Volume Pool, load balancing <br> <br> This article presents a step-by-step description for how to set up a two-server GlusterFS array. Havin...")
 
(No difference)

Latest revision as of 19:30, 7 August 2019

For Ubuntu 16.04. For Hard drive Volume Pool, load balancing



This article presents a step-by-step description for how to set up a two-server GlusterFS array.

Having two web servers behind a load balancer means that they have to synchronize the files that they serve and write to. On modern Linux distributions, GlusterFS is the easiest way to accomplish this task.

Install GlusterFS

To install GlusterFS, run the following commands on both servers:

apt-get update
apt-get install -y glusterfs-server glusterfs-client

Prepare the bricks

GlusterFS needs a file system that supports extended attributes to store its data. It creates directories in the file system and calls those directories bricks.

On a 4-GB Performance server, you have a whole extra drive that’s already partitioned for you:

[email protected]:~# ls /dev/xvde*
/dev/xvde  /dev/xvde1
  1. Format this partition with ext4 by running the following command on both servers:

    mkfs.ext4 /dev/xvde1

  2. You don’t want anything other than GlusterFS using this partition, so mount it in a hidden directory. You can put them in /srv/.bricks. If you have multiple xfs volumes per server, you can put them in /srv/.bricks1 and so on.

    mkdir /srv/.bricks echo /dev/xvde1 /srv/.bricks ext4 defaults 0 1 >> /etc/fstab mount /srv/.bricks

In the example, the term bricks is used because each directory in the setup is a GlusterFS brick. A GlusterFS volume is built of bricks (usually bricks on different hosts).

Open the firewall

Open the firewall to allow all traffic on this network. Run the following command on both servers:

ufw allow in on eth2

In this example, the network is on the device eth2. You can use the command ip addr show to see all network devices and networks associated with them.

If you added web01 to the network first, it has the IP address 192.168.0.1 and web02 has the IP address 192.168.0.2.

Link the servers

  1. Introduce the two Gluster servers to each other. The following example runs the command on web01, and tells it to link with web02:

    [email protected]:~# gluster peer probe web02 peer probe: success

  2. Run the gluster peer status command on web02 to confirm that the servers are linked:

    [email protected]:~# gluster peer status Number of Peers: 1

    Hostname: 192.168.0.1 Port: 24007 Uuid: d080d5cc-4181-4d3f-91bc-ef42bb4e8ec9 State: Peer in Cluster (Connected)

Create the GlusterFS volumes

Now you can create the volumes. Run the following command on only one of the servers:

[email protected]:~# gluster volume create www replica 2 transport tcp 192.168.0.1:/srv/.bricks/www 192.168.0.2:/srv/.bricks/www
volume create: www: success: please start the volume to access data

The parts of the command are as follows:

  • gluster - this is the Gluster command-line tool.
  • volume create - You are creating a Gluster volume.
  • www - This is the volume name. You can call it whatever you like.You will use it later in your /etc/fstab file when mounting the volume.
  • replica 2 - Every file on this volume will be replicated between 2 bricks. In this case, that means at least two servers (because there is only one brick on each server).
  • transport tcp - Use TCP/IP to synchronize the volumes.
  • 192.168.0.1:/srv/.bricks/www - This is the first brick with which the volume is built (on web01).
  • 192.168.0.2:/srv/.bricks/www - This is the second brick with which the volume is built (on web02).

For more information about these options, you can run the man gluster command.

Start and mount the volume

The volume exists, but it is not being actively synchronized nor served.

  1. Start the volume by running the following command on either server:

    [email protected]:~# gluster volume start www volume start: www: success

  2. Mount the volume in /srv/www initially. Run the following commands on both servers:

    mkdir /srv/www echo localhost:/www /srv/www glusterfs defaults,_netdev 0 0 >> /etc/fstab mount /srv/www

  3. Create the mount point, configure it in /etc/fstab, and then actually mount the GlusterFS volume.

  4. In /etc/fstab, add one special option: _netdev. This option tells Ubuntu that the filesystem resides on a device that requires network access, and to not mount it until the network has been enabled.

Test it

At this point, a file written to or read from /srv/www/* should be the same on both systems.

To test it, create a file on web01, view it on web02, delete it on web02, and then verify that it is gone on web01:

web01:

echo hello > /srv/www/test.txt

web02:

cat /srv/www/test.txt  # should print 'hello'
rm /srv/www/test.txt   # Delete the file from web02

web01:

ls /srv/www/ # Should return nothing

Move your web content to GlusterFS

In this example, /var/www must be on GlusterFS. Ensure that web01 has the correct /var/www.

If you are running this on an already live server, you must to shut down Apache on both servers. You could set up a custom down for maintenance page and health monitoring on the load balancer first if you like, but that’s beyond the scope of this article.

  1. On web01, move /var/www/ to /srv/www/:

    mv /var/www/* /srv/www/

  2. On web02, if you’re sure that you don’t need it, you can free up space by deleting /var/www:

    rm -rf /var/www mkdir /var/www

  3. Create a bindmount so that /srv/www is accessible via /var/www. Run the following commands on both servers:

    echo /srv/www /var/www none defaults,bind 0 0 >> /etc/fstab mount /var/www

You should be able to see your web content with ls /var/www on both servers.

Summary

Your /etc/fstab should look as follows on both servers:

#                
/dev/xvda1  /               ext4    errors=remount-ro,noatime,barrier=0 0       1
/dev/xvde1 /srv/.bricks ext4 defaults 0 1
localhost:/www /srv/www glusterfs defaults,_netdev 0 0
/srv/www /var/www none defaults,bind 0 0

If you run findmnt | grep srv the response should look something like this:

[email protected]:~# findmnt | tail -n3
|-/srv/.bricks         /dev/xvde1     ext4            rw,relatime,attr2,inode64,noquota
|-/srv/www             localhost:/www fuse.glusterfs rw,relatime,user_id=0,group_id=0,default_permissions,allow_other,max_read=131072
`-/var/www             localhost:/www fuse.glusterfs rw,relatime,user_id=0,group_id=0,default_permissions,allow_other,max_read=131072

This response shows that /srv/.bricks is mounted on /dev/xvde1 and that localhost:/www (the GlusterFS volume) is mounted in two places, /srv/www and /var/www (thanks to bind mounting).

GlusterFS should show everything as healthy:

[email protected]:~# gluster peer list
unrecognized word: list (position 1)
[email protected]:~# gluster peer status
Number of Peers: 1

Hostname: web02
Port: 24007
Uuid: 56e02356-d2c3-4787-ae25-6b46e867751a
State: Peer in Cluster (Connected)

[email protected]:~# gluster volume list
www

[email protected]:~# gluster volume info www
Volume Name: www
Type: Replicate
Volume ID: bf244b65-4201-4d2f-b8c0-2b11ee836d65
Status: Started
Number of Bricks: 1 x 2 = 2
Transport-type: tcp
Bricks:
Brick1: 192.168.0.1:/srv/.bricks/www
Brick2: 192.168.0.2:/srv/.bricks/www

Conclusion

You have installed GlusterFS and configured your servers to share your web content. Both servers hold a copy of the files and share changes almost instantaneously.

Where to go from here

The next article in this GlusterFS series describes how to Add and Remove GlusterFS Tutorial in a GlusterFS array.


Kategorie: Load Balancing Kategorie: GlusterFS