I worked the other way round to avoid overlap messages by excluding the
.xxx. users
:EXCLUDE.
FULLPACK *
.SPOOL. *
.PAGE. *
.DIRECT. *
DGTSRV* 0500
DGTSRV* 05FF
:END.
One can say the the .xxx. users are better as DIRMAINT gets its info from
the ALLOC info, hence the cylinder ranges are what CP uses, and not what
someone coded on MDISK cards in the $xxx$ users. But, CP protects its
SPOOL,PAGE and DRCT: it will forbid a LINK to a minidisk that overlaps them
(except for a fullpack minidisk).
Kris,
IBM Belgium, VM customer support
Berry van
Sleeuwen
<berry.vansleeuwe To
[EMAIL PROTECTED]> [email protected]
Sent by: VM/ESA cc
and z/VM
Discussions Subject
<[EMAIL PROTECTED] Horror Story. Was: DIRMAINT
.UARK.EDU> questions
10/11/05 09:38 PM
Please respond to
VM/ESA and z/VM
Discussions
Hello List,
Actually it could be dangerous to exclude the overlay check with the $*
users. I have found this the hard way today.
In the system DDR there are the $* users to overlay the system areas:
$ALLOC$, $PAGE$, $SPOOL$, $DIRECT$, $T-DISK$, $SYSWRM$ and $SYSCKP$.
DIRMAINT will issue a Q ALLOC to determin the system areas and overlay a
.<user>. on those areas. So .DRCT., .PAGE. etc. However, those areas do not
include the areas formatted as PERM. So cylinder 0 is not protected this
way. Smart readers will notice two other areas that are not protected, yes,
the checkpoint and warmstart area.
All our VM's will have a user that will be owner of all system minidisks
and
data areas. All VM's except one. This one is basically a default
installation as it came from the DDR-tape. OK, I wanted to get rid of those
annoying OVERLAP in the dirmap, so I excluded $P*, $D* and $S* not
realizing
that with the exclude on $S* it would also exclude the SYSWRM and SYSCKP
extents. And these are now not protected by the default minidisk owner and
not by DIRMAINT.
Today I moved some minidisks and the first has overwritten the checkpoint.
No problem so far (yet). The second started writing on the warmstart area.
VM died instantly. Since these areas were lost also SPOOL files are lost.
When Murphy strikes he strikes good... Of all minidisk I could overwrite I
just did the worst possible. If only the minidisks were larger than 8
cylinders, if only I had added an owner for cyl 21-27, if only, if only.
My luck, I had made a backup yesterday so I could restore the RES-pack
quite
easy but VM was out for 90 minutes in total. And it took the rest of the
day
to clean up the mess I got myself into. I guess it is time to add our
default owner user in this VM too.
Regards, Berry.
On Wed, 1 Dec 2004 16:55:30 +0100, Rob van der Heij <[EMAIL PROTECTED]>
wrote:
>There's a lot of old traditions. In a former life we would backup the
>$ALLOC$ extents with VM:Backup and even make sure the virtual address
>in $ALLOC$ matched the real UCB. Gives you a few more places to find
>your way when things go bad.
>
>Talking about old conventions, many of us are used to $PAGE$ and
>$SPOL$ as well (to reserve the spot and for documentation). I think
>they even come in the DDR system. I think later DirMaint include the
>information from Q ALLOC into the overlay checking automagically, thus
>flag the $PAGE$ as overlay. On one system where I looked they had
>fixed this by disabling the overlay checking for all $* users...
>
>Rob
>--
>Rob van der Heij rvdheij @ gmail.com
>=========================================================================
