Gluster Storage Platform

From GlusterDocumentation

Contents

[edit]

Gluster Storage Platform

Gluster Storage Platform is an open source clustered storage solution. The software is a powerful and flexible solution that simplifies the task of managing unstructured file data whether you have a few terabytes of storage or multiple petabytes. Gluster Storage Platform integrates the file system, an operating system layer, and a web-based management interface and installer. Installation takes less than 10 minutes. It supports native GlusterFS, CIFS and NFS access protocols and Infiniband, GigE, 10GigE network interconnects. You can get started with just a single system configured as a standalone NAS server, and scale from there on as needed

[edit]

What's New in v3.0 Series

Gluster Storage Platform

  • A complete storage operating system/platform built on GlusterFS cluster file system that can be easily installed on industry standard commodity hardware
  • Efficient installer that enables quick install and boot of first server, with disk formatting done post install, and supports remote install of subsequent servers
  • Integrated easy to use and feature rich web-based management UI that makes installation and configuration of large clustered storage systems quick and simple. Key management interface features are:
    • Volume manager
    • Resource manager
    • Server/disk manager
    • Log viewer
  • Built in, pre-configured support for 10GigE and Infiniband (Open Fabrics) making it easy to utilize these hi-speed server interconnects

Gluster File System Enhancements

  • Self-healing optimizations for database and virtual machine environments that require "always" open files and high availability
    • Background self-healing: enables applications to run faster as they are no longer blocked during self-healing operations
    • Checksum based healing: rsynch like healing mechanism that heals only the inconsistent blocks within a file versus the entire file
    • Healing on the fly: a file can be healed even while the file is open and application is performing active I/O to it
    • Generation numbers: Self-healing can now reliably handle tricky corner cases
  • Quick read: entire small file data is transferred to client in one atomic operation on first access to file making subsequent file accesses extremely fast
  • Pre-fetch: 'stat' information of multiple files delivered to client on access to one file and cached on client enabling faster access to subsequent files
  • I/O cache performance enhancements: ability to handle large file data cache more efficiently due to improved internal data structures
  • Performance enhancements for NFS, CIFS and webDAV client access protocols
[edit]

Download Gluster Storage Platform 3.0.3

Gluster.com

The Cube - Mirror

[edit]

Gluster Storage Platform 3.0.3 Release Notes

Release Notes for Gluster Storage Platform 3.0.3

[edit]

Minimum Requirements

  • 1GB USB Disk as Install Media
  • 1 or more storage servers (or virtual machines) with minimum configuration as follows:
    • Disk 8GB
    • Memory 1GB
    • Intel/AMD 64bit hardware
    • 1GigE or 10GigE or Infiniband
[edit]

Prepare your Install Media

Gluster Storage Platform is released in the form of a compressed USB bootable disk image. After uncompressing, you transfer the platform image on to your USB disk (without any partitions) using standard dd disk-dump utility.

Note1: Most desktops will automatically mount the USB disks upon insertion. Unmount it manually before you transfer the image.

Note2: To determine the device name of your USB disk, you can type 'dmesg | tail' inside your terminal. Alternatively you can also find from your /proc/devices file.

Assuming your USB disk is detected as /dev/sdc, you type for following commands: 

root$ wget http://ftp.gluster.com/pub/gluster/gluster-platform/3.0/3.0.4/Gluster-3.0.4.img
root$ dd if=Gluster-3.0.4.img of=/dev/sdc bs=1M
[edit]

Virtual Machine Approach

As of now we have tested on Red Hat KVM, Citrix Xen, VMware and Sun VirtualBox implementations.

[edit]

RedHat KVM

Create a 10GB disk image (server1.img): 

root$ qemu-img create -f raw  server1.img 10G

Perform the installation using USB disk image (Gluster-3.0.img): 

root$ sudo kvm -boot c -m 1024 -hda Gluster-3.0.3.img -hdc server1.img

Boot from the installed disk (server1.img): 

root$ sudo kvm -m 1024 -hdc server1.img
[edit]

USB Disk Approach

Simply insert the USB installation media you prepared from Gluster Storage Platform image and boot the system into it.

[edit]

Installation Procedure

[edit]

Install Modes

Gluster install media has two modes of installation.

[edit]

Management Console / First Server Installation

If you are starting fresh, this is your first step in the installation process. Installation will show a time-out dialog asking you to interrupt it for installation to begin. If you miss this step, it will automatically proceed to an automated remote server installation mode (explained below). Then your only option is to reboot and pay little more attention next time :). Installation is quite simple. There are only two screens to take care of, disk selection and network configuration (described below), and the whole process takes less than 10 minutes to complete.

Installation is all about preparing one server with the management console enabled. If you have just one system, you can still use Gluster Storage Platform as a standalone NAS server with its web based management GUI. Later you can add more servers and turn the system into a powerful storage cluster.

[edit]

Remote Server Installation

You require this step only when you want to add more servers to your cluster. You boot from the install media and drive the installation from the management console, available from the first server you installed, remotely.

[edit]

Disk Selection

You need to select the disk you want to install Gluster Storage Platform from a drop-down list of detected disks. The disk you select here, will be made bootable and used as the disk for the NAS backend.

[edit]

Network Configuration

You assign hostname, ip-address and gateway values. Other fields are optional. When you are done with this step, the installer formats the disk you selected and transfers the Gluster Storage Platform image on to it.

[edit]

Reboot into Management Console

Once the installation completes successfully, you will be notified about the default PASSWORD for the system. Remove the USB install media at this point and click the reboot button. It will then re-boot into a full blown NAS server + management console.

[edit]

Gluster Management Console

Once you logon after re-boot, you will be presented a management console through which you can add servers, create volumes and monitor the system. You can also manage the server entirely remotely from your browser. Simply point your browser to http://IP-ADDRESS (or hostname) of your server (i.e. the IP which you assigned during the installation). You will get the same login screen and features via the web browser as you would being physically on the installed system.

[edit]

Server Manager

Once you login, your first step is to prepare the disk you selected at the time of installation. Select the server (check box) and click prepare-data-disk. This process takes few minutes to few hours depending on how big your RAID volume is. Once this step completes, you are ready to create as many virtual volumes as you need using the volume manager.

[edit]

Add Server

The first time when you add a new server to the cluster, you will be prompted to enter an IP address range for the storage pool. You simply reserve a range of IP addresses in advance for the storage servers. For example, if you are planning to scale the storage cluster to a maximum of 20 servers, then you will enter 192.168.1.12 - 192.168.1.20, where 192.168.1.11 is the IP address you already assigned for the management console. There is no harm in assigning more than what you are planning to scale.

The next step is to use the same installation media to boot your new storage server. Do not interrupt the booting process. It will automatically proceed to remote installation method on expiry of the interrupt time-out, and return you back to the management console. Click Add Server. Your remote server will be automatically detected and you will be prompted for remote network settings.

[edit]

Volume Manager

The Volume manager is the heart of Gluster Storage Platform. If all you have is just one server, you can still turn it into a standalone NAS server. You can then add servers any time later as desired to scale the system.

By default, volume manager creates a distributed volume with a single global name space. All of the servers will appear as one large volume, when you mount the file system on your client machines. It uses an elastic hashing technique to intelligently distribute the files and to avoid hot-spots.

[edit]

Add Features

  • Mirror Feature: All your files are replicated twice. You need at least 2 serves to activate this feature.
  • Stripe Feature: Each file is spread across 4 servers to share load within a file access. It helps in performance only in HPC / Supercomputing environment where there are 100s or 1000s of clients pounding on a single large file (much larger than server's memory). You need minimum 4 servers to activate this feature.
[edit]

Protocols

  • Native GlusterFS: Native GlusterFS protocol is designed for scalability and performance. It is intelligent and efficient in file access. On the flip side, it works only on FUSE capable kernels such as Linux and Mac OS X. You need to install a tiny user-space GlusterFS client on your system in order to natively mount the volumes. It is fully POSIX compliant and requires no modification to your application code.
  • NFSv3 TCP: Distributed NFS support brings scalability to NFS without imposing any additional software requirements on the client side. Most Unix and GNU/Linux operating systems have support for NFSv3 TCP protocol.
  • CIFS: Microsoft operating systems have built-in CIFS client support, making this protocol ideal for general-purpose file storage and archival in a Windows environment.

[edit]

Resource Manager

Resource manager allows you to monitor CPU, Disk and memory graphically. We will bring in sophisticated platform management and reporting capabilities (such as IPMI, SAF-TE) in our upcoming releases.

[edit]

Summary

We have worked hard to bring you this new product offering. We hope you will download it and take it for a spin. Do write to us on the the community discussion lists about your experiences, and other feedback you may have. For reporting and tracking bugs, please follow http://bugs.gluster.com

[edit]

Frequently Asked Questions

[edit]

Where can I get an ISO image for installation, or a VMDK image?

The Gluster Storage Platform VMDK image can be downloaded from: 

 http://ftp.gluster.com/pub/gluster/gluster-platform/3.0/LATEST

At this time we do not have an ISO image available for Gluster Storage Platform. We aim to provide one soon. The USB image reference above can be used for install in the interim.

[edit]

How do I use the Gluster-3.0.3.img file I downloaded with Redhat kvm?

Right now it can be used with "virt-install" command 

 On Ubuntu or Fedora with virt-install package installed. 
 $ virt-install -n gluster -r 1024 -v --import --disk path=Gluster-3.0.3.img --disk path=disk1.img
 $ virt-viewer gluster

 Command line kvm based installation on Debian and Ubuntu
 $ kvm -hda Gluster-3.0.3.img -hdb disk1.img -m 1024 

 Command line qemu-kvm based installation on Fedora/Redhat
 $ qemu-kvm -hda Gluster-3.0.3.img -hdb disk1.img -m 1024
[edit]

How do I use the Gluster-3.0.3.vmdk file I downloaded with VMware ESX?

Instructions coming soon...

[edit]

What should I do if the master node gets corrupted?

To prevent configuration loss, we recommend users export their configuration (export/import tab) and save it. If configuration gets corrupted, the configuration can be imported to restore the machine.

To prevent data loss, in the unlikely case of a disk file system corruption on a system, we recommend users create and use mirror volumes.

[edit]

I prepared the USB image, but it doesn't boot on my machine?

Check if your BIOS supports a USB boot disk. If it does, please enable it in the BIOS for your system. Then reboot your system.

[edit]

I have configured a new volume. How do I mount it from the client side?

If you have seleted "Ethernet" and configured the volume it can be used from the client side by providing the below command. 

 $ mount -t glusterfs <MANAGEMENT_SERVER>:<VOLUMENAME-ON-WEBGUI>-tcp <MOUNTPOINT>
[edit]

My network card doesn't get detected. What should I do?

Please go through the supported hardware list, if the drivers are absent please leave a message with the hardware specs at support (at) gluster.com, we will add it with the next update.

[edit]

How do i mount nfs and cifs shared volumes?

Please go to client machine where you are allowed access through Access control. Specify mount as below for each type of reexport. 

mount -t nfs <ip-management-server:/nfs/<volumename-webui> <MOUNTPOINT>

mount -t cifs -ousername=<adusername>,password=<adpassword> //<ip-management-server/<volumename-webui> <MOUNTPOINT>

From Windows it could be accessed by from "Network Places" with \\<ip-management-serveR>\<volumename-webui>

[edit]

How do i mount a volume through glusterfs native client?

Please go to client machine where you are allowed access through Access Control. Specify mount command as below on client 

mount -t glusterfs <ip-management-server>:<volumename-webgui>-<transport> <MOUNTPOINT>

Here <transport> is the transport layer you have your volume exported. Required values cane one of these ie. "tcp" or "ib-verbs"

[edit]

My RAID controller is not getting detected. What should I do?

Please go through the supported hardware list[1], if the drivers are absent please leave a message with the hardware specs at support (at) gluster.com, we will add it with the next update release.

[1] Please check Fedora 11 hardware list.

[edit]

I have the first server ready. How do I add new servers?

See the Add Server section earlier in this document (manual section).

[edit]

A node I configured has issues. How do I remove it from the servers list?

This functionality will be delivered in an upcoming release.

[edit]

The format disks process is hung. What should I do now?

While formatting all the disks at once will work, we recommend formatting disks one at a time. Formatting too many disks at once can slow down the system to a crawl making it unresponsive.

[edit]

Can I have ssh login to the node?

Console access to the platform is disabled on purpose. Modifying the operating system image directly may make the system unstable. Particularly the subscription network updates may overwrite or get confused, when it detects manual changes. The data partition is always accessible to the user in standard EXT4 format (backward compatible with Ext2/3). In case of recovery or migration, users can bypass the Gluster Platform by using any standard GNU/Linux rescue CD. If you still prefer command-line access and full control over the system, GlusterFS directly on standard GNU/Linux or Solaris operating systems is the right choice for you.

Retrieved from "http://www.gluster.com/community/documentation/index.php/Gluster_Storage_Platform"

 

Copyright © 2009 Gluster, Inc. All Rights Reserved.