It is possible to share a disk r/o among multiple images. We're doing it with 14 SUSE SLES7 images sharing a common /usr disk under VM. This is the 294 disk in our setup.
There are several "gotcha's", though. When you finish writing to the disk from the one image allowed to do so, you should issue these commands from that image: mount /usr -o ro,remountsync;syncshutdown -h now Then this image should NEVER link the disk r/w again. ALL OTHER IMAGES should link it r/o. If using VM, put LINK MASTER 294 294 RR in the image's directory. (where 294 is the disk to link r/o) Also, each image should use this in the parmfile, or /etc/zipl.conf, or whatever to ensure that the 294 is to be used r/o by Linux: ... dasd=293,292,294(ro),295-29F ... and /etc/fstab should specify that the disk is mounted "ro". What we do is have TWO /usr disks, a 194 and a 294. One (294) is linked r/o by all the running images, and the other (194) is linked r/w as 294 by the test system. When we need to make the new stuff on the 194 production, we go through the remount/sync above, shut down ALL the linux images, and use the VM directory facility to change the 194 to 294 and 294 to 194, swapping the disks, and bring everything up again. Been working fine for months. "Only two things are infinite: the Universe and human stupidity, and I'm not so sure about the Universe." - Albert Einstein Gordon Wolfe, Ph.D. (425)865-5940 VM Technical Services, The Boeing Company > ---------- > From: McKown, John > Reply To: Linux on 390 Port > Sent: Friday, August 16, 2002 5:58 AM > To: [EMAIL PROTECTED] > Subject: Re: Can Linux disks on VM be shared read-only? > > I should think that if all Linux instances accessed the disk as R/O, you > will not have any problems. However, if one Linux has it mounted in Write > mode, then you could run into logical inconsistancies. That's because > Linux, > like all UNIXes, heavily buffers disk I/O and defers writes. This means > that > meta-data (directory entries and inodes) could be written before the > actual > data is written. If another image tries to read, then the data it reads is > garbage. As I recall AIX had a GFS (Global File System) which could do > this. > And if Linux ever implements "raw" disks, this might be possible. Make > sure > that the MDISK is LINKed as RR and that the fstab mounts the disk as RO. > > -- > John McKown > Senior Technical Specialist > UICI Insurance Center > Applications & Solutions Team > +1.817.255.3225 > > > -----Original Message----- > > From: Nix, Robert P. [mailto:[EMAIL PROTECTED]] > > Sent: Thursday, August 15, 2002 3:46 PM > > To: [EMAIL PROTECTED] > > Subject: Can Linux disks on VM be shared read-only? > > > > > > I'm trying to break up a RedHat Linux image so that some of > > the major pieces of software reside on their own disks, with > > the thought of sharing them as read-only minidisks between > > several images. I had a problem during the install, and have > > called the IBM support center about it, but in the process of > > discussing the problem, when I told him my ultimate > > intentions, his response was Oh, you can't do that. It's not > > supported; you have to use NFS for that." > > > > This really took me by surprise, because, unless Linux just > > flat won't work with a read-only disk, I'd think that'd be > > the way to share things... I mean,... It's VM, right? Why > > incur the overhead for NFS when you could just read the > > minidisk directly. And one image wouldn't really even know > > the other image was sharing the device... > > > > Am I way off base? > > > > ---- > > Robert P. Nix internet: [EMAIL PROTECTED] > > Mayo Clinic phone: 507-284-0844 > > 200 1st St. SW page: 507-255-3450 > > Rochester, MN 55905 > > ---- > > "In theory, theory and practice are the same, > > but in practice, theory and practice are different." > > > >
