Juraj,
I tried to distinguish two major cases: defined and scratch
A. defined volumes:
a0. "def v <stgpool> <vol>" - STGNEW record is made in volhistory,
numwrites is 0
a1/1. tape is written first time - numwrites becomes 1
a1/2. tape is reclaimed - becomes "Pending" and later "Empty"
a2/1. tape is written second time - STGNEW record *is not* made in
volhistory, a STGREUSE record is made instead, numwrites becomes 2
a3...aN. tape was written N-1 times and now is written for N-th
time - number of STGREUSE records is N-1, numwrites=N
aN+1. "del v <vol>" - STGDELETE record appears in volhistory, numwrites
counter is deleted/reset
Result: numwrites is reliable as far as volumes are not deleted but number
of STGNEW (1) and STGREUSE (N-1) records keeps the information how many
times the tape was written
B. scratch volumes
b1/1. volume is auto-defined in the pool and gets written immediately -
STGNEW record is made in volhistory, numwrites becomes 1 (maybe you can
catch the millisecond while it is 0 :-)
b1/2. the volume is reclaimed - becomes "Pending" - numwrites is still 1
b1/3. the volume is auto-deleted from the pool - STGDELETE record is
inserted in volhistory, numwrites is cleared
b2/1. same *libvolume* is auto-defined in same or another pool - new
STGNEW record is made in volhistory, numwrites becomes again 1
...
Result: numwrites is always 1 but number of STGNEW records tracks how many
times the tape was written.
Conclusion: slight improvement of the select covers all combinations:
select volume_name,count(*) as "Usage count" from volhistory where
type='STGNEW' or type='STGREUSE' group by volume_name
The information is reliable again with the disclaimer volhistory was not
deleted with type=all/stgnew/stgreuse.
Zlatko Krastev
IT Consultant
Salak Juraj <j.salak AT ASAMER DOT AT>
Sent by: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
29.04.2003 17:06
Please respond to "ADSM: Dist Stor Manager"
To: ADSM-L AT VM.MARIST DOT EDU
cc:
Subject: AW: Can the volume write pass value be used to track
media life?
Hi Zlatko,
I tried your select statement,
it returns number of times a tape has been defined as NEW,
but not overall write passes.
Seems like one would have to write a script which would maintain two
values:
1) summ of write passes for each tape and
2) "write pass" as delivered by Q VOL.
Once a tape becomes either pending or reused the number from (2) should
be
added to (1) and reset to zero.
I guess this only would count write pasess while I believe
the sony´s definition of 30.000 uses
counts read-only passes as well??
Maybe an even simpler solution
would be to store Q ACTL regulary
in external files and sum all "MOUNTED" lines for each tape (grep, awk).
This could count read-only mounts as well,
or even write and readonly mounts separately.
Another point to consider is mount/unmount pass,
I am not sure for how many "uses" it counts.
I guess during both tape mounts and unmounts the label is read thus
counting for 2 read passes?
If yes, this would reduce the real tape uses to 30.000/2=15.000 TSM
mounts.
Another point,
re-labeling tapes would cause losing the track for particular tapes.
Questions, questions..
regards
juraj Salak
-----Ursprüngliche Nachricht-----
Von: Zlatko Krastev/ACIT [mailto:acit AT ATTGLOBAL DOT NET]
Gesendet: Dienstag, 29. April 2003 14:37
An: ADSM-L AT VM.MARIST DOT EDU
Betreff: Re: Can the volume write pass value be used to track media
life?
I looked through the thread and see no comprehensive answer so will try:
write pass shows number of times the tape gets written. You have two
options:
1. Defined volumes - once reclaimed (or "move data", "move nodedata" used)
the volume becomes "Empty" and next round pass begins. For defined volumes
you can see write pass number be 0, 1, 2, ....
When the volume is deleted from the pool this number is reset - it is
tracked for volumes only but not for libvolumes! Thus the number is
reliable if the volumes are never deleted from the pools.
2. Scratch volumes - upon reclamation the volume is deleted from the pool.
Thus the number is reset after each use and you cannot have the number
over 1. The number also cannot be 0 because the volume is defined to the
pool immediately before first write. Thus for the scratch volumes the
number is always 1.
Usually scratch volumes are used and "write pass" number is useless. But
this information can be tracked other way:
select volume_name,count(*) as "Usage count" from volhistory where
type='STGNEW' group by volume_name
This would work correct only if volume history is not deleted for STGNEW,
i.e. "del volh t=all" is not used.
Zlatko Krastev
IT Consultant
Laura Buckley <buckley AT STORSOL DOT COM>
Sent by: "ADSM: Dist Stor Manager" <ADSM-L AT VM.MARIST DOT EDU>
17.04.2003 07:45
Please respond to buckley
To: ADSM-L AT VM.MARIST DOT EDU
cc:
Subject: Can the volume write pass value be used to track
media life?
Hi,
I have a customer who must prove to a governing agency that they are
adequately tracking media life. They are using AIT-3 media which SONY
says has a lifecycle of 30,000 uses and a shelf life of approximately 30
years. So - in order to comply with their regulations, they must show
that they are tracking the uses. Does the value in the field called
Write Pass Number in the volume record equate to what the manufacturer
considers "uses"? I believe this to mean a complete write pass from
beginning to end of media.
TIA
Laura Buckley
STORServer, Inc.
buckley AT storserver DOT com
|