You guys rock. Thank you so much for the information. I'll visit the site you recommended and do some reading. Maybe I can even mock this up in a Xen VM pool.
Cheers,
Clay
On Mon, 2007-06-04 at 23:54 +0200, Schaefer Christian wrote:
Hello Clay, > My only concern is the level of complexity that > this would introduce into our environment. Dealing with cluster infrastructures is never easy. Compared with the NFS server topology you currently use, the webservers will then access a local GFS filesystem for operating system purposes and several other GFS filesystems for common data (web application files or media files). Due to the fact, that each flesystem can be mounted on each of the nodes - provided each of the filesystems has enough journals - you just have to publish data to one single filesystem to be delivered to customers visiting your websites. In other words, the complexity added on the one side lasts in more simplification on the other side. > There also doesn't seem to be much documentation on building > diskless Linux clusters that I've found. Please referr to http://sourceforge.net/project/showfiles.php?group_id=156679. There you can find preconfigured boot images and the com.oonics cluster suite, which provides administrative tools to build and maintain a shared root cluster. Documentation is little, but once i have seen a HOWTO to build a shared root cluster node from the bootimage. Please send mail to ATIX, I am shure they have the document laying around somewhere... ;-) > Are you using specialized tools for the configuration? Usually we use the standard tools provided by a standard linux installation. Moreover the gfs_tool binary or other programs included in the GFS package assist while handling system maintenance tasks. Greetings, Christian Christian Schäfer Abt. IT-Applications Messe München Tel.: +49 (0) 89 949 21985 E-Mail: christian.schaefer2@xxxxxxxxxxxxxxxxx WWW: www.messe-muenchen.de > -----Ursprüngliche Nachricht----- > Von: linux-cluster-bounces@xxxxxxxxxx > [mailto:linux-cluster-bounces@xxxxxxxxxx] Im Auftrag von > Clayton Dillard > Gesendet: Montag, 4. Juni 2007 18:53 > An: linux clustering > Betreff: Re: AW: Red Hat Cluster & GFS for a LAMP > > Christian, > Thank you for the notes and detail. This sounds > promising. My only concern is the level of complexity that > this would introduce into our environment. > > There also doesn't seem to be much documentation on building > diskless Linux clusters that I've found. > > Are you using specialized tools for the configuration? > > Clay > > > On Mon, 2007-06-04 at 18:23 +0200, Schaefer Christian wrote: > > > Hallo Clayton, > > > Does anyone here run such a configuration and does it > work well? > Definitely yes. Munich trade fair center ist hosting > the fair's websites itself. Due to availability and > performance issues we have decided in 2005 to renew our > server pool and migrate it to the diskless shared root > concept to to achieve the aim, that by administrating only > one server you could manage all servers. The topology > mentioned consists of 8 PHP frontend servers responsing on > the customers' requests if they were invoked by a pair of > LVS/Keepaliced loadbalancers and a couple of backend servers > and database clusters. MySQL is currently used in an > actice/passive style, but is operated on an two node diskless > shared root cluster as well. > > > > I'm wondering if anyone on this list is currently using RH > > Cluster Services and GFS for a LAMP environment and if so, > > are there any pitfalls to watch out for, or any tips that I > > should know before getting started on the engineering phase. > As clusters are more complex than single host systems, > a shared root environment is on step more complex. Hence you > should precisely plan your environment to run, because there > are more components involved, and some piece of software are > not designed to run on such cluster environment. Especially > PHP must be patched due to some session locking behavior, so > that the webservers run more stable. Another example is > writing logfiles. Here the applications run on the cluster > should be aware of the GFS file locking mechanism. Using a > single application log file on all of the cluster nodes can > be a performance killer. > > > > I was reading an article this weekend on the RH site about > > using a shared (GFS) root partition on SAN and booting up > > cluster nodes to this shared root, the premise being that > > patches and config changes are done once, to the clustered > > root on all servers, as opposed to each server separately. > This environment is developed and operated for the > munich trade fair center by ATIX (http://www.atix.de). If > you'd like to have more information about our environment, > just reply to my message. > > > Greetings, > > Christian Schäfer > Abt. IT-Applications > Messe München > > Tel.: +49 (0) 89 949 21985 > E-Mail: christian.schaefer2@xxxxxxxxxxxxxxxxx > WWW: www.messe-muenchen.de > > > > > -----Ursprüngliche Nachricht----- > > Von: linux-cluster-bounces@xxxxxxxxxx > > [mailto:linux-cluster-bounces@xxxxxxxxxx] Im Auftrag von > > Clayton Dillard > > Gesendet: Montag, 4. Juni 2007 17:53 > > An: linux clustering > > Betreff: Red Hat Cluster & GFS for a LAMP > > > > Hi Folks, > > I'm hoping for some advice here. We have a hosting > > service where we run a LAMP stack. Currently our environment > > is comprised of a pair of Apache servers which are load > > balanced via a pair of hardware balancers, and a pair of > > MySQL servers that are in a Master/Slave configuration. The > > two Apache servers use a common NFS export as shared storage > > and the DB servers just use local disks. > > > > I've been reading up on RH cluster services and GFS, and from > > what I see, it sounds like a great solution for us, > > especially since we are bringing on more and more customers > > and thus have higher loads and availability requirements. > > > > I'm wondering if anyone on this list is currently using RH > > Cluster Services and GFS for a LAMP environment and if so, > > are there any pitfalls to watch out for, or any tips that I > > should know before getting started on the engineering phase. > > > > I was reading an article this weekend on the RH site about > > using a shared (GFS) root partition on SAN and booting up > > cluster nodes to this shared root, the premise being that > > patches and config changes are done once, to the clustered > > root on all servers, as opposed to each server separately. > > Does anyone here run such a configuration and does it > work well? > > > > Also, I've not been able to find any good books that are > > focused mostly on Red Hat clustering and GFS... Does anyone > > know of any good books on this subject? > > > > Thank you, > > > > -- > > Clayton Dillard <cdillard@xxxxxxxxxxxxxxxxx> > > RPS Technology, LLC > > > > -- > Linux-cluster mailing list > Linux-cluster@xxxxxxxxxx > https://www.redhat.com/mailman/listinfo/linux-cluster > > -- > Clayton Dillard <cdillard@xxxxxxxxxxxxxxxxx> > RPS Technology, LLC > -- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
|
-- Clayton Dillard <cdillard@xxxxxxxxxxxxxxxxx> RPS Technology, LLC |
-- Linux-cluster mailing list Linux-cluster@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/linux-cluster
