Gluster FAQ
From GlusterDocumentation
Contents |
General FAQ
What is Gluster?
Gluster is a GNU cluster distribution aimed at commoditizing Supercomputing and Superstorage. Core of the Gluster provides a platform for developing clustering applications tailored for a specific tasks such as HPC Clustering, Storage Clustering, Enterprise Provisioning, Database Clustering etc. Gluster is designed for massive scalability and performance from ground up. Makes heavy use of muti-casting and RDMA where ever possible.
Minimum system requirements to use Gluster?
To use Gluster, you will need
- A cluster of ia32 or x86-64 systems with atleast 512MB RAM (1GB recommended) each.
- A supported GNU/Linux distribution pre-installed (for GlusterHPC) on the master node.
Does Gluster support my distribution?
Currently Gluster has been tested on
- Debian
- Slackware
- Redhat Enterprise (and clones such as ScientificLinux and CentOS)
- Fedora Core (1-4)
- Ubuntu
- Gentoo
- SLES (Suse Linux Enterprise Server) 10
- OpenSUSE
NOTE:
The following are known issues and are being worked on:
- Distributions should be installed without LVM/Software Raid. Support for LVM/SW Raid is being added currently.
- Fedora Core 5 with SELinux turned on fails to root login on client nodes.
Why is Gluster a non-GNU project?
non-GNU means `not yet GNU'. For a project to be certified as a GNU project, it has to be submitted to the GNU evaluation team, which on approval will be dubbed an official GNU project. Gluster will soon be submitted for GNU approval. For more information on this topic, read from here http://www.gnu.org/help/evaluation.html
Bulde meaning?
'bulde' is a slang word used for word 'burude' (from kannada language) , which means brain/head in English.
'bulde' is also nick name of one of the gluster developer @ #gluster (on irc.gnu.org).
Gluster Versioning - More details about Gluster Versioning.
How can I contribute?
Contribution can come in may forms. Word of mouth marketing, documenting and writing articles, bug reports, patches.. If you are developer, you can start with Gluster Hacking Guide.
How portable is Gluster?
Ideally, Gluster is portable to any architecture which has a GNU toolchain and a bootloader which can be built with the GNU toolchain for that architecture.
Gluster currently works only on the GNU/Linux operating system.
Gluster should work on any GNU/Linux distribution, which is file system standards compliant.
Can I add my question here?
If you do not find your question answered here and if you think it is a frequently asked question, you may add your question here. One of us will fill in the answer.
Can I edit these wiki pages?
Anonymous editing of this wiki has been suspended after seeing wide spread vadalism on this site. But you can contribute by creating a user account of your own choice. That way by making the best usage of this documentation and help us with feedbacks, corrections
GlusterHPC FAQ
How is GlusterHPC better than ROCKS or OSCAR?
GlusterHPC is an application of Gluster for High Performance Computing. It packages a rich set of clustering software stack (see GlusterHPC Applications).
GlusterHPC is
- Designed for massive scalability (16 nodes or 65,000 nodes makes no difference). Much of the building blocks of Gluster are already powering worlds top supercomputers.
- Portability (across distributions and architectures).
- Modular and extensible.
- Built on Gluster Platform which extends clustering technology beyond HPC to database, storage, enterprise provisioning, etc.
- Very easy to use with a clean dialog based front-end.
- Backed by supercomputing experts.
- Supports multi-casting and Infiniband.
- Centralized remote screen control.
- Very easy to add new features or customize.
- Doesn't require a database server to store configuration information.
My /etc/hosts file got overwritten by GlusterHPC. There was no back up. What to do?
Now, GlusterHPC supports backing up the original file (v1.0-rc3 onwards) . /etc/hosts file in the Image will be stored as /etc/hosts.orig-gluster and now the new entries are appended to the original file.
Similarly, network config files, /root/.ssh dir will be backed up as <filename>.orig-gluster
GlusterEP FAQ
How is GlusterEP better than SystemImager or RedHat Kickstart?
Both SystemImager and Kickstart are very hard to setup and manage. Gluster is better because
- Supports highest number of distributions - Debian, Slackware, Redhat and clones such as ScientificLinux (CentOS), Fedora Core (1-4), and Ubuntu.
- Automated image capture: Just remote boot your system and select capture image option.
- Server administration console: User and image management (view/add/delete)
- Multi-user support: Each user can organize his/her own set of images.
- Secure: Images are password protected and transfers are encrypted.
- Easier to setup: Just boot the server and answer few questions.
- Multiple NIC support.
- Backup: Can be used as server/workstation network backup software for entire OS.
- Dual-boot images are supported.
- Extensive hardware compatibility.
- Images are compressed for efficient storage.
My system does not support PXE. Can I use GlusterEP?
Gluster provides two alternatives for situations where the BIOS does not support PXE boot.
- client boot disk - You can create a client boot disk on a USB stick or in the form of a bootable ISO (which can be burnt to an empty CD-ROM/RW).
- pxe boot disk - You can create a PXE boot floppy/USB/ISO which will boot your system as though it supported PXE from the BIOS itself.
NOTE: client boot disk is the only option when GlusterEP client is outside the master node's network and hence cannot perform PXE boot even with a pxe boot disk.
Developers FAQ
How can I contact the developers?
The developers can be reached on the mailing list gluster-devel@nongnu.org. Visit http://www.gluster.org/mailing-list.php for subscription.
You can also reach the developers on IRC at irc.gnu.org (freenode) in the #gluster channel. Visit http://www.gluster.org/core-team.php for nicknames.
Why Gluster has its own tool chain?
Most of Gluster works in LIVE mode and Gluster is distro independent. So everything is built from source. To ensure sanity of the build involving everything, right from the kernel to the shell. Having our own toolchain makes the overall system predictable like compatible compiler version and libc version, specific compiler features and language support and makes Gluster build-system independent (and hence, distro independent)
It also makes easier to port Gluster to new architecture, only a matter of ensuring that Gluster compiles in the new toolchain used.
How to build a new tool chain?
Gluster toolchains are built using crosstool. Crosstool is available here http://kegel.com/crosstool/crosstool-0.42.tar.gz . Steps for building toolchain for gluster involve:
- Extract crosstool
- Edit options file, <arch>-static.dat (i686-static.dat for i686 arch) and change TARGET to 'i686-gluster-linux-gnu', make sure that USE_SYSROOT=1.
- Edit demo-<arch>.sh file and set the RESULT_TOP, TARBALLS_DIR to relavent paths.
- Edit demo-<arch>.sh and set GCC_LANGUAGES to "c, c++, fortran".
- Edit demo-<arch>.sh and uncomment the line relevent to newest gcc and glibc.
Eg. uncomment the line with following command in it, to build gcc-4.1.0 and glibc-2.3.6.
eval `cat i686-static.dat gcc-4.1.0-glibc-2.3.6.dat` sh all.sh --notest
- Now run run the demo-<arch>.sh script to build the toolchain.
I am interested in porting Gluster to XXX arch, Where do I start?
Gluster is designed to be portable from scratch. Except boot-loader, entire Gluster is built from source using Gluster's own tool chain. Your first step would be to build a tool chain for your platform and adding your boot loader binary to the repository. Rest is mostly compilation work. Look for more information in Gluster Hacking Guide.
