Mount disks

[cromwoli]

Hi, I am running VirtualBox V6.1 on a Linux Mint host system. I have created a OpenVMS VM from the VMDK file provided. It starts up and runs VMS after I boot DKA0: In the VM machine I have created, I added not only the VMDK but an additional disk which hoped would also be present when I booted VMS up Sadly I don't see anything other than DKA0: How do I mount the other disk as part of the system startup? I have added photographs illustrating how the VM is set up in VIrtualBox, and also what I see when I execute the SH DEV command in VMS.
Any ideas?

[cct]

You need to initialise it, then mount it - usually in SYSTARTUP_VMS.COM

See the help for INIT as the default values are not sufficient.

I used the folowing dor a 20Gb data drive:

Code: Select all

$ INIT DKA100: data1 /system /DATA_CHECK=write /DIRECTORIES=5000        -
        /EXTENSION=8 /HEADERS=500000 /NOHIGHWATER /DATA_CHECK=WRITE     -
         /STRUCTURE=5 /VOLUME_CHARACTERISTICS=HARDLINKS

and added to SYSTARTUP_VMS.COM: $ mount/syste DKA100: data1 /noass

[sms]

Code: Select all

> [...] Sadly I don't see anything other than DKA0: [...]

   "see" _how_, exactly?

   _I_ see "V922$DKA100:" in your SHOW DEVICE report.  (If you're
interested in disks only, "show device d".)

> You need to initialise it, then mount it [...]

   INITIALIZE once, MOUNT at system start.

> [...] usually in SYSTARTUP_VMS.COM

   I'd suggest creating a separate DCL script to do the MOUNT (for all
your disks other than the boot volume, SYS$SYSDEVICE), and then running
("@") that script from SYSTARTUP_VMS.COM.  I find that in the long run,
the less editing of SYSTARTUP_VMS.COM, the better.

   In a cluster, you might also want to SET DEVICE /SERVED at system
start (before MOUNT).

[cromwoli]

Thanks guys. I will follow that advice & post the results.
S.

[arne_v]

I don't mind putting thing in SYSTARTUP_VMS.COM (after I have cleaned it up).

But note that if the disk contains page/swap files then it need to be mounted earlier in SYPAGSWPFILES.COM.

[dgordon]

There are no swap files on x86.

If you are paging to a non-system disk then it's probably also worthwhile to create a dump file on that disk and use DOSD (Dump Off System Disk), which for x86 is easily implemented with the new SET DUMP command.
Example:
$ SET DUMP/DEVICE=(DKA100:) ! x86 only
You may not dump to any shadowed non-system disk, nor may you specify a volume that is a member of a shadow set. Dump to a shadowed system disk works as in previous architectures. The dump is written to the shadow master member.

Unlike on previous architectures, you can modify the DOSD device without an intervening reboot and you can change your mind any time while the system is up. SET DUMP/DEVICE will even take a list, but this is less useful than one would think because the Dump Kernel will attempt to mount the devices in list order and will use the first one it finds.

If the target disk is mounted, SET DUMP will verify that a dump file can be found in [SYSn.SYSEXE] or issue a message if the device is not currently mounted, but the command will succeed. (SYSn is almost always SYS0 for x86 and if it isn't, you'd better know what you're doing. If you're uncertain what system root is in use on a particular system, SHOW LOGICAL SYS$TOPSYS will show you.)

$ SHOW DUMP will list the devices for DOSD and the saving of the error log buffers. If the device(s) are mounted, it will display the file specification and size for each file.

[arne_v]

No swap file.

Does that mean that it swaps to page file or does it mean no swapping at all (no xxxO states)?

[pocketprobe]

There is no support for swapping on x86. Page 22 on this doc.
https://vmssoftware.com/docs/VSI-X86V91-RN.pdf

[dgordon]

IIRC (and I might not because in port time, this happened a long time ago and I'm not the responsible engineer), the only state left is COMO and it's lasts exactly between process creation and the first time the process is scheduled.

[arne_v]

Interesting.

But I guess it makes sense.

Even 35 years ago the logic was: you have swapping => buy more memory

And while back then memory tuning was all about trimming processes to fit in available RAM, then today memory tuning is about increasing various parameter to make it possible for processes to use the available RAM.

So ripping out no longer needed codet to simplify things makes sense.

Bye LEFO and HIBO (and real COMO) - you will not be missed!