James English wrote:
> We already turned on file logging with LOGTOOL to capture logging event
> number 105 (NM file close). We found one job that is deleting the files, but
> we suspect there may be more than one process involved.

How about also including log event 160 CM file close?

> One idea I had was to change the file permissions on the MCOMIN files so
> that purging the file is not allowed. Read, write, lock, etc. would be
> allowed, but purging would not. So if something tries to purge the files,
> then the purge would be disallowed and the job or program would abort, and
> we'd know exactly what is purging the files.
> 
> Is there a way to change the file permissions to disallow purging of a file?
> Here is the LISTF,-3 on one of the MCOMIN files (with creator and lockword
> blanked out):
> 
> ********************
> FILE: MCOMIN.MDATAB20.PROD         
> 
> FILE CODE : 0                   FOPTIONS: BINARY,FIXED,NOCCTL,STD
> BLK FACTOR: 128                 CREATOR : XXXX.XXXXXXXX        
> REC SIZE: 2(BYTES)              LOCKWORD: XXXX                 
> BLK SIZE: 256(BYTES)            SECURITY--READ    : ANY        
> EXT SIZE: 4(SECT)                         WRITE   : ANY        
> NUM REC: 257                              APPEND  : ANY        
> NUM SEC: 16                               LOCK    : ANY        
> NUM EXT: 1                                EXECUTE : ANY        
> MAX REC: 257                            **SECURITY IS ON       

I do not believe MPE provides that feature to disallow purging. I tried by
building a file, removing WRITE access on the file (ALTFILE ;ACCESS=), but
could still purge it.

Its been a very long time since I did MANMAN (version 3),
you could attempt to take WRITE access off, since most processes only
read MCOMIN. But I can't remember if some items need write, like
sequence numbers or such.
(Back then (198x) we eventually moved our MCOMIN,OCOMIN,PCOMIN,GCOMIN files
  into the database -- no more flatfile)

I have an FCLOSE trap for an XL that traps closing file 0; with the
options to work as is, report error, ignore the call; controlled
via an envvar. It should be easily changed to detect deletes of MCOMIN.
A CM version for the SL could also be crafted.

Keven

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *