| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 5220.1 | check the errorlog | VMSSG::JENKINS | Kevin M Jenkins VMS Support Engineering | Wed Jan 29 1997 10:45 | 40 |
|
Well that is sort of a tough question... I'll make some assumptions.
We'll assume the Host Clear is a result of the DUDRIVER timing out
an IO...
Check the errorlog for MSCP Command Timeouts
What happens is that there is a timeout interval specified by the
controller... for an HSJ it's 200 seconds. There is a timer routine
that wakes up every 200 seconds and checks the CDDB for the HSJ if
it finds there is an IO outstanding it issues a GET COMMAND status
to the controller it expects the controller to respond with a command
status indicating it's making progress... if it does then the timer
routine is happy. If it doesn't the driver will issue a host clear..
Now if there were no IO on the CDDB we issue a Get Unit Status to
unit zero... all we want to do is let the controller know we still
exist..
When the timer routine is doing this it sets a flag called IMPEND
if when it starts it finds this flag is still set we will issue
a host clear and log an MSCP Immediate Mode command timeout... this
is an indication that the controller has simply stopped talking to us.
Now if you get the latest DUDRIVER kit for your VMS version and
CPU type.. VAX of ALPHA. The DUDRIVER has enhanced errorloging
for the COMMAND Timeout where we log the actual command that
has timed out, this can often be helpful in figuring out what
is going on...
Third.. if you already have the latest code... it's timer code
is designed to recognize an HSJ/HSD that is lying about making
progress on a command and shoot it anyway... this used to result
in a hang.
So you need to check the version of you DUDRIVER and check your
errorlog for evidence of why the host clear was issued.
Kevin
|
| 5220.2 | Thanks | KEIKI::WHITE | MIN(2�,FWIW) | Wed Jan 29 1997 13:42 | 9 |
|
Kevin,
How many versions of OpenVMS did they retrofit these changes
for DUDRIVER into, Also I have seen alot of host clears from Tape
drivers to HSC's. Are the timeouts the same or am I mistaken, Do
we even issue host clears from the magetape drivers to HSJ/HSC's?
Bill
|
| 5220.3 | | TKTVFS::AEBA | _Mamoru Aeba MCS/CHIBA/JAPAN | Thu Jan 30 1997 02:46 | 17 |
| Hello,Kevin
Thanks for your comment.
I understood your comment below. Correct?
1. DUDRIVER's timer routine wake up every 200 second(for HSJ's).
2. Check the CDDB for HSJ.
3. If outstanding command is available,then it issues "GET COMMAND
STATUS"
4. If HSJ not respond, DUDRIVER set the IMPEND flag.
5. Next timer routine wake up again,then if IMPEND flag still set,
it issues controller reset.
Thanks.
/Mamoru
|
| 5220.4 | | VMSSG::JENKINS | Kevin M Jenkins VMS Support Engineering | Fri Jan 31 1997 07:49 | 34 |
|
.1
We backported the DUDRIVER changes through V5.5-2
The TUDRIVER has pretty much the same timer code but we didn't
do anything with it.
.2
< 1. DUDRIVER's timer routine wake up every 200 second(for HSJ's).
correct the actual timer value is controller specific.
< 2. Check the CDDB for HSJ.
< 3. If outstanding command is available,then it issues "GET COMMAND
< STATUS"
correct
< 4. If HSJ not respond, DUDRIVER set the IMPEND flag.
sort of... while the GET COMMAND STATUS is outstanding the IMPEND
bit remains set... so if 200 seconds later it is still set it
is assumed that the controller is hung and we shoot it. We also
shoot it if we don't like the command progress status that is
returned for the GET COMMAND STATUS
> 5. Next timer routine wake up again,then if IMPEND flag still set,
> it issues controller reset.
Right.
Kevin
|
| 5220.5 | Thanks. | TKTVFS::AEBA | _Mamoru Aeba MCS/CHIBA/JAPAN | Sun Feb 02 1997 03:06 | 0
|