From mboxrd@z Thu Jan 1 00:00:00 1970 From: Patrick Caulfield Subject: Re: [linux-lvm] Could vgscan read a file of "preferred" LV name -> minor device id ? Message-Id: <20020111104116.GB1274@tykepenguin.com> References: MIME-Version: 1.0 Content-Disposition: inline In-Reply-To: Sender: linux-lvm-admin@sistina.com Errors-To: linux-lvm-admin@sistina.com Reply-To: linux-lvm@sistina.com List-Help: List-Post: List-Subscribe: , List-Unsubscribe: , List-Archive: Date: Fri Jan 11 04:42:01 2002 List-Id: Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit To: linux-lvm@sistina.com On Fri, Jan 11, 2002 at 07:11:00AM +0000, Piete Brooks wrote: > I'm getting complaints from NFS users that when servers reboot, they often > find that their mounts have the wrong data. I've never really worked out > which end hold what in NFS FHs, but the basic idea is that a client identifies > a FH which encodes (major dev id,minor dev id,inode number,version). > This appears to break down when a new LV is created ane the server rebooted. > As the /dev// nodes are numbered strictly in the order that they are > discovered, all LV minor device IDs on subsequent VGs are renumbered, so any > remote caching of FHs breaks (( as with SCSI )). > > Is that about right ? > > Would it be possible for vgscan to read a "hints" file consisting of / > and minor device number pairs ? > > As each / is found, if it's in the file, the nominated minor device id > is used; if it's not found, then the next minor device id which is not listed > in the file is used. > > [[ I use devfs, so using existing data in /dev/ would not help -- but I think > it would do no harm to use that data by default ]] > I thought Heinz had fixed this in 1.0.1, what version are you using (tools & kernel)? patrick