Hello Anne, Just some things that popped into my mind.
>VMTEST 191 and looks like this: >USER WVLNX30 WVLNX30 512M 1024M G ..snip.. >DEDICATE 0A81 163C >DEDICATE 067E 165F > >VMTEST's USER DIRECT on first level: ..snip.. > MDISK 163C 3390 000 10017 LX163C MR > MDISK 165F 3390 000 3338 LX165F MR The guest VM has a minidisk of 3338 cylinders. Shouldn't that be 3339 cylinders? A 3390-3 has 3339 cylinders. I would guess to have either 001 3338 as a minidisk or 000 3339 for a full pack disk. Secondly, in this case you dedicate the 165F to the linux machine, so fro m cylinder 0 through 3338. Wouldn't that overwrite cylinder 0, thus replaci ng the CP volume information with whatever linux would like to see? Our zLin ux guests format the volume with a label something like 0x0200. Should that label be placed on cylinder 0 the hosting VM (either host or second level VM) we wouldn't be able to find LX165F afterwards. And the last point, perhaps not applicable, beware of running 3rd-level guests. They don't perform as well as second level guests. The CPU overhe ad in 3rd level is very large due to the fact that SIE assist doesn't work a t that level. (Assuming you run in LPAR mode, which is the only way on z-series today) We had an issue when we ran VSE guests on a guest VM. It turned out the VSE had on overhead ratio 1:2 or worse because all CPU had to be emulated by the host. The VSE couldn't use SIE assist in this configuration. SIE assist works only two levels deep, LPAR mode being one of them. Regards, Berry.
