On 9/6/06, Wakser, David <[EMAIL PROTECTED]> wrote:
So, if the shared disk does NOT start ay Cyl 0 but at Cyl 1, the MWV
is worthless and I cannot share disks?
If you want the VSE guests in this LPAR to use reserve/release, you
need MWV to have CP simulate it (that's the *virtual* reserve/release
part). The cyl0 is only when it needs to go to the real device so
other LPARs can participate.
I think what I need to understand is: who says CP has to issue any
reserve/release? Does not VSE (via the lock file or some internal locking
mechanism) handle this? Why does CP need to be involved?
The control unit is supposed to acknowledge R/R to provide exclusive
access to the volume for more than one channel program. If you would
have guest A issue the reserve and then get guest B from the same z/VM
image try it, the control unit would grant access because the LPAR
already holds a reserve, so the control unit thinks this is fine. What
you want is CP to be arbiter and issue the real reserve for the volume
as long as one of the guests holds a reserve.
BTW, there are no LPARS involved here. And the reason why not all
shared packs start on cyl 0 is because we wanted to retain uniqueness of
DASD volids whenever possible. Thus, the VSE DASD starts at REAL CYL.
If it's just 1 z/VM LPAR then you have no need for real reserve to be
issued because there is nobody outside who can try. And afaik VSE only
does the R/R for some disks (like the lock file) and for the other
disks there is logic (in the lock file) that protects the datasets
(rather than reserve the entire volume anytime you want to write a
shared dataset).
I hope I did not make it more confusing that it has to be. Rob
--
Rob van der Heij
Velocity Software, Inc
http://velocitysoftware.com/
