On Thu, 21 Sep 1995 08:04:09 EDT, Melinda Varian <[email protected]> wrote:
> I have an on-going problem with the tape mount exits on my VM server.
> If the exit logs off too soon after the tape is attached to DSMSERV,
> ADSM decides there is something wrong and issues this message:
>
> ANR5241E Exit Machine DSMTAP1 not logged on, mount for A15689 cancelled.
>
> It issues this message and cancels the mount even though the tape has
> already been attached to it.
>
> On the other hand, if the exit doesn't log off quickly enough after the
> tape is attached to DSMSERV, DSMSERV tries to use the exit machine again,
> finds that it is logged on, decides there is something wrong with it,
> and stops using it. Ultimately all of the exits are disabled.
>
> And the timing here is very close. I've been experimenting with
> whether my exit should wait either 7 or 8 seconds after the tape
> drive has been attached to DSMSERV before it logs off. I seem to have
> optimised the exit now in the sense that I get about equal numbers of the
> two failures, but this is getting to be a pain. I don't see the point of
> ANR5241E. Why should ADSM care whether the exit machine has logged off
> by the time it notices that the tape has been attached? (Or is it perhaps
> checking the state of the exit before it notices the ATTACH? I am quite
> certain that the ATTACH has completed before my exit logs off.) Is there
> any way around this problem? (I am running VM/ESA 1.0 with level 14 of
> the server.)
>
> Melinda Varian,
> Princeton University
>
Melinda,
Let me try to explain how this is supposed to work. But first, you shouldn't
code wait in your mount exit after attaching the tape. As you have seen,
that can result in an exit machine being "offlined". APAR PN64299 was put
out quite some time ago (you already have it) because exit machines were being
offlined too quickly in cases like this. Your exit machine should simply
attach the device and log off.
The logic on the server goes something like this: if the server detects a
device at the requested vaddr, the mount is considered "complete"; else if the
exit machine is not logged on _AND_ the device isn't attached, the mount is
considered to be "cancelled". This is where message 5241 comes from. Note
that the virtual address is always checked first. In fact, before message 5241
is issued, the vaddr is checked TWICE (both before and after the query of the
exit machine). Message 5241 says, in effect, "this mount has been cancelled
because the device wasn't attached and the exit machine assigned to the mount
isn't logged on".
This sounds as if VM is processing the logoff before the attach. I remember
that some flavor of VM used to perform asynchronous DETACH processing, but I
thought ATTACH was always synchronous. I highly recommend removing the
wait from your DSMMOUNT EXEC after doing the ATTACH - just logoff immediately.
If the problem persists, send me a note offline and I'll look into it.
Wayne Hineman
IBM Almaden Research Center
|
