[Info-vax] Logging image activations

[Jan-Erik Söderholm]

P. Sture wrote:
> In article <NGQFm.12425$U5.180157 at newsb.telia.net>,
>  Jan-Erik Söderholm <jan-erik.soderholm at telia.com> wrote:
> 
>> FrankS wrote:
>>> On Oct 27, 6:26 pm, Jan-Erik Söderholm <jan-erik.soderh... at telia.com>
>>> wrote:
>>>> Hm, maybe some ACL on the files to enable auditing of
>>>> "execute" access ? That would be easy to add to the files
>>>> anyway, I'm not sure about how easy it is to get reports.
>>>> Maybe some mangling of the ANA/AUD output would do...
>>>>
>>> That option would at least allow you to be more selective about which
>>> executables get logged, instead of all image activations.  The
>>> assumption is that you know in advance the executables of interest.
>> Yes, the images of interest are all in 4-5 applications
>> specific directories. And they are also all on a application
>> specific device. Most/all images that I'm not interested in
>> are on the system device.
>>
>> I'll probably experiment with both the SET ACC/ENAB=IMAGE method
>> and some ACLs to enable access auditing and see what's easiest.
>>
> 
> I generally recommend that SET ACCOUNTING/ENABLE=IMAGE is reserved for 
> diagnostic purposes only.
> 
> For example, DCL SET commands invoke SET.EXE and these get logged in the 
> accounting file, so as you can imagine image accounting fills the disk 
> up pretty quickly. A DCL procedure doing repetitive processing can 
> easily run up tens of thousands of image accounting records.
> 
> Unless you have your accounting files on a different disk from the audit 
> files, you are risking grinding your system to a halt.
> 

Yes, absolutely. That is the main drawback of genarally
enableling image accounting. I will probaby do some test
using the porposed "$ INSTALL CREATE imagename /ACCOUNTING"
by Jeremy Begg. I guess that that will introduce some (one?)
extra steps (INST REPL ?) when copying a new image from the
development/test system. But besides of that, it seems as a
nice addition.

One other issue is that some of the "images" I'm interested in
are build as a large EXE with different entry points, and if
possible I'd be interested in which of the entry points gets called.
They are setup with the name of the EXE and the entry point
in the menu system used. But that is maybe a little harder to do
without some loogging in the code itself (I guess).

Jan-Erik.