Everhart, Glenn (FUSA) From: Glenn C. Everhart [Everhart@gce.com] Sent: Friday, May 07, 1999 9:16 PM To: Info-VAX@Mvb.Saic.Com Subject: Re: [Fwd: RE: OpenVMS Shadowing Disk Geometry Restrictions Lifted] I've done this, in essence, with the vddriver that's on the Fall 1998 sigtapes (just released). (I put "canonical" versions for vax and alpha on the tapes since there's been a lot of confusion about that.). Look a bit further in my drivers of course and you'll also find various flavors of shadowing drivers, journalling drivers and the like. The questions about remapping /dev/devicename to devicename: remind me of some hacks done in DOS-11. Suppose I give a /dev/devicename filename to RMS. What gets down to the XQP? I note that / and \ are illegal even in ods-5, which will make specifying such beasts rather a pain. However, if we presume some layer of software (a C runtime maybe) converts it to [dev]devicename, then I would expect RMS to "think" of this as a normal filespec. If we have a (zero length) FILE [dev]devicename.; on the disk, and create a softlink to devicename: where the device named is NOT file structured, an io$_access might "sort of" work. (Safety does softlinks for single files, but didn't really contemplate this.) What Safety does is reassigns the channel to the desired target device (adjusting counts) and sends the io$_access on to the desired target. This might possibly be usable, at least to the extent that the user's channel now points at the correct device even if it isn't filestructured. Give it a try. It isn't logicals but it may be usable. The code that creates softlinks may need tweaking though, for this case; I'll look into that. I've been fiddling with the softlink code in the past few days with the idea of arranging that a directory softlink will allow one to access a directory AND its contents on another disk without needing softlinks on all files explicitly. The key is to notice when an open is seen for a file with DID pointing to the linked to directory but where the channel points at the old device, and updating the channel. In addition to the DID matching a directory that has been softlinked from (and the new dir still open), the UCB of the file needs to match the UCB the directory was linked FROM for this to make sense. Hopefully the result of this fiddling will be that the softlink implementation will allow whole directory trees to be moved across volumes transparently. It seems more general than device logicals somehow... Glenn Everhart everhart@gce.com Greg Thomas wrote: > > On Thu, 06 May 1999 21:18:06 -0400, Everhart wrote > in article <37323F4E.D81DF0F0@gce.com>: > > >It makes perfect sense to even not require size the same; just make > >the shadowset size the minimum of the component sizes. > > Ideally, though, you could shadow (e.g.) a 2Gb and 4Gb disk together > (to give 2Gb of space), remove the 2Gb disk, remove the 2Gb limit on > the 4Gb disk, and voila - zero down-time disk capacity increases. > > Greg > -- > This post represents the views of the author and does > not necessarily accurately represent the views of BT.