Hi Daniel, You could do that setup, but a better approach would be to make the server processes on storage servers directly talk to clients on the webservers. (instead of re-exporting) just as shown in the example link you gave. Regards Krishna On Thu, Mar 20, 2008 at 8:59 PM, Daniel Maher <dma+gluster@xxxxxxxxx> wrote: > On Thu, 20 Mar 2008 20:11:16 +0530 "Krishna Srinivas" > <krishna@xxxxxxxxxxxxx> wrote: > > > It can suggest option "a" if the glusterfs client is run either on > > storage1.example.com or storage2.example.com > > > > It can suggest option "b" if the glusterfs client is run on a > > different machine. > > > > How you interpret that example depends on where you run the client. > > Thank you for the prompt response, Krishna. > > Suppose i have four webservers and two back-end storage servers. I > would like to have both of storage servers be mirrors of each other, > and in turn, i would like to have the webservers be able to interact > with the storage "cluster" (such as it is) in a read/write capacity. > > Using Gluster, would the proper approach here be to set up both of the > storage servers as gluster servers / clients of each other (option "a"), > then export the mirror as a volume to my webservers (running the client > only) ? In this fashion i would hope to achieve a position whereby one > of the storage servers could suddenly burn to the ground, but > operations would not be affected, since the volume remains accessible. > > As per the "simple high availability storage" wiki page, this would > seem to be the case - but i'm more than happy to hear any other > thoughts on the matter. > > Thank you all. > > > > -- > > > Daniel Maher <dma AT witbe.net> > > > _______________________________________________ > Gluster-devel mailing list > Gluster-devel@xxxxxxxxxx > http://lists.nongnu.org/mailman/listinfo/gluster-devel >
