Well, it might just boil down to trying to do 5 stripes when I had only a single volume--I just tried again today with no striping and the setup worked. I could swear I tried that yesterday, but perhaps I just redid definitions on the IPLed install-linux system (hit abort after I got the LVM error, redefined the LVM filesystems, and retried the install) instead of starting from scratch like I did now.
The reason we want to do striping is so that linux can take advantage of the static PAV aliases when we do get our mod-54s. I'm trying to build something now that we could clone to the mod54s later. However, I was already intending to set up the "application" LVM group with at least 5 minidisks (even if they were all from the same physical mod54). The root filesystem isn't really big enough to justify creating a lot of minidisks for it. I think I was striping the LVM just as a matter-of-course (stripe one LVM, stripe them all), not really thinking about the mechanics of it. Think I'm sorted now--thanks for all the help! -----Original Message----- From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of RPN01 Sent: Friday, March 14, 2008 9:35 AM To: LINUX-390@VM.MARIST.EDU Subject: Re: Is it possible to do an initial SLES9 install on a partially-LVM root? One disk or twenty, LVM for all or just a few filesystems, it should still work. So you still have some problem in your definitions that hasn't been revealed here yet. Since you're running into the problem during the install, there aren't a lot of files you could post here for review. But could you walk through, in detail, what you are doing during the install, from the point of formatting the DASD through to the error? What options are you selecting on which panels, and what steps you go through to define each physical volume, volume group and logical volume? Perhaps, given this information, others here can spot the problem, or test your config on their own systems and find the mis-step. If you can, take "pictures" of the major windows involved in the process. I'm not sure that the group here will allow posting them, but you could, at the very least, create a Flickr account and post them there, and then supply the URL to get to them here. -- Robert P. Nix Mayo Foundation .~. RO-OE-5-55 200 First Street SW /V\ 507-284-0844 Rochester, MN 55905 /( )\ ----- ^^-^^ "In theory, theory and practice are the same, but in practice, theory and practice are different." On 3/14/08 8:15 AM, "Collinson.Shannon" <[EMAIL PROTECTED]> wrote: > We don't have mod54's yet--this is just a mod9, and we'll have static > PAV aliases defined for the mod54s--but I assume the same logic applies. > Does this mean that I can't use LVM for the four mountpoints (/usr, > /var, /opt and /home) at all until we have PAV set up? I could define > minidisks on different volumes, but at the moment we're > storage-crunched--I've only got 3 mod9s to work with. If I stick with > the current plan (/ on a separate disk, the other 4 on LVM), I'd have at > least two filesystems overlapping--3 volumes but 5 mountpoints. I don't > have a problem with the I/O requests waiting in that case, but it > appears that the SLES9 install is trying to mount the LVM filesystems at > once, since otherwise I would have thought I'd just have a "slow > install", not get the "file descriptor 3 left open" error. Everything > worked fine when I was installing all of root on a single minidisk... > > -----Original Message----- > From: Linux on 390 Port [mailto:[EMAIL PROTECTED] On Behalf Of > Bruce Hayden > Sent: Friday, March 14, 2008 9:01 AM > To: LINUX-390@VM.MARIST.EDU > Subject: Re: Is it possible to do an initial SLES9 install on a > partially-LVM root? > > You need to remember that you can't start more than 1 I/O to an > address (real or virtual!) at a time. This is a hardware restriction. > This is why PAV was invented - so that the PAV alias addresses can be > used to start parallel I/Os to the same base volume. So - if you > don't have PAV, then it will not help to split up your mod 54 into > minidisks. Linux may try to start more than 1 I/O at a time (since it > can start one to each virtual address) but z/VM will only execute > them one at a time and the other I/O requests will wait. > > If you had HyperPAV, then splitting up the mod 54 into minidisks > works, because then z/VM will use the PAV pool for that device to do > I/Os to that device in parallel. If you have static PAV, the same > thing also works, but the alias addresses have to be set up in > advance. If you had SLES 10, then instead, I'd recommend that you > define virtual PAV aliases to the guest with 1 minidisk and use > multipath on Linux for overlapped I/O. But - this is not supported on > SLES 9. > > > > LEGAL DISCLAIMER > The information transmitted is intended solely for the individual or entity to > which it is addressed and may contain confidential and/or privileged material. > Any review, retransmission, dissemination or other use of or taking action in > reliance upon this information by persons or entities other than the intended > recipient is prohibited. If you have received this email in error please > contact the sender and delete the material from any computer. > > SunTrust and Seeing beyond money are federally registered service marks of > SunTrust Banks, Inc. > [ST:XCL] > > > > > > ---------------------------------------------------------------------- > For LINUX-390 subscribe / signoff / archive access instructions, > send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit > http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390 ---------------------------------------------------------------------- For LINUX-390 subscribe / signoff / archive access instructions, send email to [EMAIL PROTECTED] with the message: INFO LINUX-390 or visit http://www.marist.edu/htbin/wlvindex?LINUX-390
