Microsoft Mail v3.0 (MAPI 1.0 Transport) IPM.Microsoft Mail.Note
From: FOSTER, Simon, SADS
To: INTERNET:ADSM-L AT VM.MARIST DOT EDU/
Subject: RE: Cleaning drives in a SCSI robot
Date: 1996-10-17 16:55
Priority: 3
Message ID: 3EEA2FABF427D011B0C440000006A509
Conversation ID: RE: Cleaning drives in a SCSI robot
------------------------------------------------------------------------------
How about scheduling an AUDIT LIBRARY command? We found that works - at
How about scheduling an AUDIT LIBRARY command? We found that works - at
least with the smaller tape units - because every tape (including the
cleaning tape) gets mounted. I admit it's inelegant (the last thing you
want to do when your drive is dirty is run every single tape through it)
but it seems to be our only option...
until this splendid suggestion for DEFINE CLEANTAPE is implemented.
Simon Foster
System Services
Distributed Systems
Technical Operations
Level A/3
National Westminster Bank PLC
Goodmans Fields
74 Alie Street
London E1 8HL
Phone: (+44/0) 171 480 2168 <ITS: 73-2168> <Mobile: (+44/0)
802 187593>
Fax: (+44/0) 171 480 2848 <ITS: 73-2848>
GOAP: Y001SF3
Internet: simon.foster AT natwestuk.co DOT uk
----------
From: INTERNET:owner-adsm-l AT VM.MARIST DOT EDU/
Sent: 17 October 1996 09:04
To: INTERNET:ADSM-L AT VM.MARIST DOT EDU/
Subject: Re: Cleaning drives in a SCSI robot
On Wed, 16 Oct 1996 09:39:35 -0400 Price, Robert R. said:
>Has anyone determined a way to automatically schedule a tape cleaning
>cycle in a SCSI robot?
It would be REALLY great if ADSM were enhanced to provide this function,
as it is very difficult to do outside of ADSM. What we have done is to
write a Rexx exec that runs every so often (5 minutes) to determine which
tape drives are currently in use. It keeps track, by tape drive, how
many
minutes each drive has been in use. It makes the assumption that if a
drive was in use when it looked, that it was in use for the entire 5
minutes. On average, this is should be a safe assumption. Then, when
a threshold is exceeded (i.e., for 8mm, we clean our drives after every
25 hours of head use), it sends a message to our operations staff
to clean the tape drive. It continues to send the message every 30
minutes, until the drive is cleaned. The operations staff runs a
procedure to actually clean the drive. The procedure is written in
Rexx, but it only walks them through a manual procedure to put in the
cleaning cartridge. However, it also resets the drive use counter
to zero since the drive has been cleaned - this stops the messages from
getting sent to them. Oh, one other thing - the message would not get
sent to the operator if ADSM was actively using the tape drive; it would
only send the message if the drive was not in use. I.e., it waits until
the drive is available before asking the operator to do something.
We use this for our 8mm robot (Lago Datawheel), and it has helped.
The next step that I had wanted to do, but have not had time to do, was
to actually automate the cleaning process. On the Datawheel, there is
not only a scsi interface to the robot, but also a serial interface. We
were going to use the latter to control the robot to clean the drive(s).
This ended up being a rather large design effort, as we wanted to make
sure that ADSM would not attempt to mount a tape in the drive at the same
time that it was being cleaned. We never got around to doing this part.
We also have since gotten a Q47 DLT robot. While we also track drive
usage for the tape drives in this robot, we do not know how often they
need to be cleaned. We have heard that DLT drives do not need to be
cleaned very often at all (very unlike the 8mm drives). We have had
these drives for 5 months now, with fairly active usage, and have not
cleaned them once. The cleaning light has not come on for any of the
drives either.
For IBM: What I would like to see ADSM do is the following:
1) For each tape drive, keep track of the number of minutes (or hours)
that the drive is in active use. If you can determine actual head
use, so much the better. Display this info in the output of
QUERY DRIVE FORMAT=DETAILED.
2) Add a feature to allow tape drive cleaning:
a) Allow us to specify the number of hours of head use which will
trigger
a drive cleaning. Threshold should be specifyable on either a
drive
basis, or at least device type basis.
b) Allow us to specify the slot (element number?) that a cleaning
tape is stored in. Also allow us to specify the number of times
that this cleaning cartridge can be used. E.g.:
DEFINE CLEANTAPE LIBRARY=libname ELEMENT=nn USES_LEFT=nn
c) Also provide a QUERY CLEANTAPE command.
d) When a drive cleaning is triggered, do the following:
- If no cleaning tape has been specified for the library, simply
issue a message to the activity log indicating that the
drive/head
use counter has exceeded the drive's threshold.
If a cleaning tape HAS been specified (i.e., by DEFINE CLEANTAPE)
then:
- wait until drive is not in active use
- quiesce activity for that drive (lock it?)
- mount the cleaning cartridge, wait 2 minutes, put it back
- resume activity for that drive (unlock it?)
- reset the drive/head usage statistic to zero
- decrement the counter for number of uses that the cleaning
tape has left.
- if the cleaning tape counter reaches zero, issue a message
to the activity log requesting that the cleaning cartridge be
replaced.
This would be a BIG help to us. Currently about the only thing our
operations staff does with our ADSM server is to clean the drives. If
we could automate this as well, it would be a big help.
I understand that some robots, particularly 8mm robots, have capabilities
to clean themselves. The LAGO Datawheel has this capability. There are
2 problems with this. First, the 8mm drives themselves do not indicate
that they need to be cleaned early enough. We have found that by the
time
the drive thinks it needs to be cleaned, we have already seen I/O errors
from ADSM on the drives. Active cleaning can forestall buildup of a
residue which degrades the signal quality, and can be difficult to get
rid
of. The second problem is that when the robot is cleaning itself, ADSM
may
attempt to use the same drive that is being cleaned. There is no
coordination between the robot and ADSM. This can result is a wedged
robot.
..Paul
Paul Zarnowski Phone: 607/255-4757
Cornell Information Technologies Fax: 607/255-6523
Cornell University US Mail: 315 CCC, Ithaca, NY
14853-2601
|
