Thanks, I'll follow your suggestions and post my experience on the list. =
Answering Aneesh, I have four regular 750 MHz Pentium III's, 256MB RAM, =
40G disk. For the moment though, I intend to only use one m/c for SSI =
> From: Brian J. Watson=20
> "Aneesh Kumar K.V" wrote:
> > What hardware configuration do you have. I will suggest=20
> to go for SSI
> > over UML.=20
> You'll be able to bring up a single node UML cluster with the
> instructions and tarball I put out a couple of weeks ago. When you try
> to bring up a second UML node, however, things go bad. John Byrne is
> trying to figure out what's going wrong. Once he's got it worked out,
> I'll release 0.6.1. My SSI Clustering Using UML HOWTO should be ready
> about that time, as well.
> > If you find any problem and you hack the cluster_mkinitrd so
> > that it will generate the /linuxrc that will mount root =20
> fs over nfs.
> Start by hacking the /linuxrc in a ramdisk generated by
> cluster_mkinitrd. You do this by gunzipping it, loopback mounting it,
> editing /linuxrc, unmounting, and gzipping it again with the -9
> parameter. Once you figure out how hack the /linuxrc script,=20
> it's fairly
> easy to hack a --nfs=3D<nfs_server> option into cluster_mkinitrd that
> generates your modified initrd.
> As I said in previous e-mails, NFS has sloppy coherency guarantees.
> Migrating processes can get confused if they see a different=20
> version of
> an open file (migration is supposed to be transparent to the process).
> > If you really want to try some minimum configuration then=20
> you can even
> > have it without a shared storage. In that case you have to=20
> make sure you
> > have both the nodes hardisk in sync.( I run here the last=20
> > ).
> Process migration is not terribly useful in this environment,=20
> but other
> aspects of SSI clustering (e.g., distributed process relationships,
> clusterwide fifos, clusterized devfs, etc.) work quite well without a
> shared root.
> > Is there any chance that we will have CFS in the next release. ?
> Version 0.6.1 will have working UML support and Laura's distributed
> /proc work. CFS won't be ready for that release.
> Pretty soon, I will have the CVS repositories up on SourceForge. Dave
> thinks he'll be checking in a limping version of CFS sometime in the
> next month, although initially it may just work as a non-root file
> system. Once it works okay as a root file system, I'll=20
> probably roll the
> 0.7.0 release.
> Brian Watson | "Now I don't know, but I been told it's
> Linux Kernel Developer | hard to run with the weight of gold,
> Open SSI Clustering Project | Other hand I heard it said, it's
> Compaq Computer Corp | just as hard with the weight of lead."
> Los Angeles, CA | -Robert Hunter, 1970