Paul, I thought it was all my fault, but after reading your issues, perhaps not.
I had a dozen servers that needed a disk added into LVM (sles 10 - kernel 2.6.16.60-0.42.5). I used a scripting tool to send the appropriate dasdfmt,fdasd,pvcreate,vgextend. Of those 12 identical servers, 2 had issues. It was reporting that a particular uuid could not be found when doing the vgscan. I dared not reboot. Issuing pvdisplays, pvscan, and lvdisplays I figured out that all of the disks were there except the new one. The new one had a completely different uuid than what vgscan was complaining about. So, I ended up issuing pvcreate --uuid against the new disk and then (I think) a vgcfgrestore and all was well. I figured somehow my vgextend command might have picked up the wrong disk, but I couldn't find any other that had that uuid in it. I chalked it up to gremlins. I'm sure I didn't blow any limit since this was 6th disk added to the LVM. Since you can duplicate, I think reporting it to your support provider might be called for. Marcy "This message may contain confidential and/or privileged information. If you are not the addressee or authorized to receive this for the addressee, you must not use, copy, disclose, or take any action based on this message or any information herein. If you have received this message in error, please advise the sender immediately by reply e-mail and delete this message. Thank you for your cooperation." -----Original Message----- From: The IBM z/VM Operating System [mailto:[email protected]] On Behalf Of Paul Raulerson Sent: Friday, October 30, 2009 3:16 AM To: [email protected] Subject: Re: [IBMVM] z/LInux, LVM, and minor disasters. Hi Mark - I would have agreed 100% with you up until a few days ago. I am still not sure it was LVMs fault. There were a LOT of DASD volumes attached to this instance, and I think I may have blown some limit somewhere. Part of the reason there were multiple LVMs is that I had a lot of 3390-3's around, and LVM, at least at one time, seemed to have trouble with creating volumes that were much larger than 100gigs using 3390-3s. Fell into a historical habit I suppose. It worries me enough, and the fact I can duplicate it, makes me worried enough to not want to use LVMs for a while. I was kinda hoping someone else had ran into this, but perhaps it is more likely I am just doing something wrong -Paul On Oct 30, 2009, at 1:27 AM, Mark Post wrote: >>>> On 10/30/2009 at 1:23 AM, Paul Raulerson <[email protected]> >>>> wrote: > -snip- >> Has anyone else ran into this? > > Not without some error messages and the like to go on. > > LVM doesn't care about device names, or DASD address ordering, etc. > All it cares about is if it can find the UUIDs it expects for all > its PVs. Those are written on the disk volumes when pvcreate and > vgextend is done. > > If it can't find all those UUIDs, then it will throw a fit. Usually > that means that a PV was added to the VG, and that volume was not > available at the next reboot. Most often that's the result of not > re-running mkinitrd and zipl. (YaST will do that for you > automatically if you use it, otherwise you have to remember to do > it. Not sure if Red Hat has a similar mechanism to keep things in > synch.) > > > Mark Post >
