ctrl+t meaning of items?

[Fredrik.eriksson]

Hi,

The topic might not be perfect, but couldn't figure out a better way of naming it.

I have rather large copy going out to a tape drive and I kind of forgot to add a /log to the command.

So my question is, the IO=NUMBER is probably IO operations to the drive, is there anyway to calculate how many blocks has been written to the drive using IO number?
NODE2::MYUSER - 750D 10:46:00 COPY CPU=00:01:02.46 PF=728 IO=820914 MEM=152

Best regards
Fredrik Eriksson

[Joseph Huber_1 · Accepted solution]

As far I see COPY is not telling values for the current operation (unlike BACKUP, which outputs the number of blocks to the output saveset), just the DCL summary for IO: this is the total number of IO operations since login.
So if the copy operation just started after login, then IO is the number for both input and output of the copy operation. And since it is, depending on the devices, probably buffered IO, it is not precisely the number of blocks.
So only COPY/LOG will give You the number of blocks, but not during copy, only after each file copied.

[Robert Gezelter]

Fredrik,

I believe that you will find that the values displayed by CNTRL-T are the same values that you will get from SHOW SYSTEM. The IO Count is the sum of Buffered and Direct IO Counts.

The components of the IO count can be displayed using SHOW PROCESS/ACCOUNTING. To look at tape drive activity, SHOW DEVICE/FULL will show the cumulative number of operations on the device.

- Bob Gezelter, http://www.rlgsc.com

[Fredrik.eriksson]

Not the answer I was hoping for but the one I thought I'd get.

I figured as much that it was counting BIO and DIO, and afaik there is not way of figuring out how many blocks were written during 1 IO count.

Best regards
Fredrik Eriksson

[Hoff]

COPY to tape? That wasn't intended to report its progress, and it's not been known locally to be particularly reliable around tapes, and it doesn't deal with reel changes or various media errors.

BACKUP is my preference here.

The block size with COPY is determined by the MOUNT and the INITIALIZE.

If you need a notification here, then add a simply command (blind) into the type-ahead buffer. \mail nla0: myuser\ is a reasonable choice, as it's easy to type blind. If I'm the only one logged in on the box, then \reply/all/bell done\ or some such also works nicely.

[Hein van den Heuvel]

>> The topic might not be perfect, but couldn't figure out a better way of naming it.

How about naming it after the true objective of the topic? "What it the IO size used by COPY"
You might want to clarify that by adding "To tape".

You can then start the topic with "I was hoping to use CONTROL-T on an OpenVMS x.x system to obtain this, but no joy".

On OpenVMS 8.3 a real control-t to copy will report, in addition to the IO counts report something like:

Copying: (X% completed) Y blocks copied of Z

( a 'fake' out-of-process control-T, using dcl$ctrlt_pid only report the classic cpu/io line. )

>>> afaik there is not way of figuring out how many blocks were written during 1 IO count.

On OpenVMS 8.3 COPY /BLOCK will allows you to just TELL copy what to uses, beside informing you of the default (124 blocks).


On older OpenVMS version, I typically used ANAL/SYSTEM ... SET PROC test ... SHOW PROC /CHAN .... SHOW PROC/RMS=(FAB,RAB).
The USZ/RSZ fields (in the extended rab for 8.3) will re-confirms the size used.
The RFA will give in indicator where in the file(s) copy is working.
PROCIO is also 'nice' for this purpose.

Another, crude but effective way?
1) Just before, or early on during copy:
$ SET CACHE/RESET ! resets top level counts, not block data
2) Just after, or towards the end of your patience:
$ SHOW MEM/CACHE/FULL
3) Check the histogram for popular IO sizes.

hth,
Hein.