OK, I think I see where you are going with this... I have done what I believe
you want to do... try this...
Set the HIGHMIG value of the first storage pool to 100
(this says don't migrate)
but if you don't have a next pool defined when the one fills backups stop.
Sooo....
Set the NEXTPOOL to what ever you want... then when the first one fills the
backup activity goes to the nextpool but migrations don't start... then to clear
the first pool to where ever you can alter the nextpool and drop the highmig.
Here is how I've used this...
DISKPOOLPTR 0 space allocated, 100% high, 100% low, nextpool=diskpool1
DISKPOOL1 ### space allocated, 100% high, 100% low, nextpool=diskpool2
DISKPOOL2 ### space allocated, 100% high, 100% low, nextpool=diskpool3
DISKPOOL3 ### space allocated, 80% high, 20% low, nextpool=tapepool
TAPEPOOL points off to a 3494 ATL with 3590's
I've read all the behavior stuff in one of the manuals somewhere but that was
then and this is now and I have no idea which manual is it in....(or red book)
All diskpool1's volumes are 2 GB files where each file is on a unique physical
volume, diskpool2's volumes are 2 GB files across the same set of physical
volumes, diskpool3's volumes are 0.5 GB files across the same set of physical
volumes. All the volumes are 4.5 GB SSA disks. This is my pseudo striping...
the diskpool space is about 210 GB's I was seeing when I had all files on all
volumes available to adsm at the same time there was some I/O wait time
present... after doing this (only letting ADSM write to each individual physical
volume at a time) the I/O wait time reported by "iostat" under AIX went to less
than 0 ! 0.#... way low... which is good but it is a pain in the rear to
migrate stuff... OH and how I handle things is point all your copygroups to the
DISKPOOLPTR, then when you want to migrate diskpool1 just point DISKPOOLPTR to
DISKPOOL2, flip DISKPOOL1's nextpool to where you want to migrate it (tapepool)
and alter the HIGH & LOW mig values... when it migrates you just reset
everything... use this same logic to clear diskpool2 & 3...
It is real slick... I can't really say if this has improved the backup times
because this specific node is the newest server and isn't taking in what the
others are but it really made a big difference in the I/O wait time on the disks
as mentioned above...
Hope this helps...
later
Dwight
______________________________ Reply Separator _________________________________
Subject: Re: ADSTAR Problems
Author: ADSM-L (ADSM-L AT VM.MARIST DOT EDU) at unix,mime
Date: 1/22/97 8:29 AM
On Mon, 20 Jan 1997, Giovanni NAPOLI wrote:
> I have the following resources:
> - 1 ADSM Server for AIX-RS/6000 - Version 2, Release 1, Level 0.6/0.6
> - 1 EXABYTE 10e
> - 1 EXABYTE 10h
>
> I have defined 2 Management Class and 2 Storage Pool one for each
> EXABYTE JUKEBOX respectively.
>
> Is there a way to configure ADSTAR so that:
> - from the client I just specify the command
> dsmc schedule -password=password
> and (when the schedule start for that node) ADSTAR backup all files
> into the first Storage Pool and when it is full automatically ADSTAR
> use the other Storage Pool ???
>
> - if it is possible I don't want to use the migration and
> I don't want to use the DIRMc option in the dsm.sys unix
> client node.
>
> HELP ME PLEASE ... THANK'S A LOT
>
Mon, 20 Jan 1997 Barry Fruchtman wrote:
> Giovanni,
> Use the NEXT=stgpool parameter on the DEFINE or UPDATE STGPOOL
> commands and you'll get the spillover, where stgpool is the
> storage pool name associated with the second library.
Sorry if I avail of this occasion ....
Maybe that my problem exposition was not clear ...
But I have checked the sintax of DEFINE STGPOOL command for sequential
access storage pools and I have not found the description for
the NEXT parameter.
However you can specify a value for the NEXTSTGPOOL parameter
(expample NEXTSTGPOOL=second) in the definition of the first
stgpool, which means a data migration starts from the first storage pool
to the second storage pool when the occupancy of first storage pool is
more than HIGHMIG value and the migration continue until the percentage
of the first storage pool occupancy is LOWMIG value.
With the configuration above-mentioned ADSTAR cannot put,
automatically, the data into the second storage pool when the first is
full, but simply moves some data from the first storage pool to the
second and it continues (with other new backups) to put data into the
first.
If it is possible, from a client, I wish to backup data into the ADSTAR
storage pools and when one of them is full ADSTAR automatically uses
another one without data migration. How can I do that ???
Thanks a lot for your attention
|
