The next backup will delete all versions from TSM.
--
Phillip
(901)320-4462
(901)320-4856 FAX
(901)652-6337 Cell
(901)384-6337 Home
-Original Message-
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED] On Behalf Of
Avy Wong
Sent: Friday, April 13, 2007 04:17 PM
To: [EMAIL
This has been discussed before. If TSM mirroring and the error is a tsm error
like the application failed, then only the tsm mirror that was being written to
is corrupted. When the application comes back up it can detect and handle
this. If it was hardware mirror only then the error will be
We have a script and it does a move data on any tape (primary or copy)
that is older then (not read or written in) x days. We are currently
using 6 months as our x. This makes sure that offsite tapes do not sit
for 7 years also.
--
Phillip
(901)320-4462
(901)320-4856 FAX
-Original
You can do a move volume and it should move all the good files. You may
have to do this a couple of times. This should get all the good files
off the tape. Now try the restore. If that wont work then the only
thing is delete the volume with discard=yes and you have lost data. If
it is current
First time I heard that rule was in a Robert Heinlein book call The Moon is a
Harsh Mistress. They even had an acronym for it TANSTAAFL there ain't no
such thing as a free lunch.
--
Phillip
(901)320-4462
(901)320-4856 FAX
-Original Message-
From: Salak Juraj [mailto:[EMAIL
We use two card or one. We have both TSM names in the DNS tsm_primary
and tsm_backup. We then use the name we want to use in the dsm.opt
file. The backup is a private network we use just for backups. We do
not route to or from this network. Thus we can be sure that a server is
not going out
I believe there is still a reason to mirror at the TSM level. If you mirror
at the hardware level only and TSM writes a corrupt entry to the DB or log,
you are hosed since the hardware mirror will write it to both copies. If
TSM is doing the mirror, it can see that it had a problem writing to
I have noticed this. I did not know if it started with the 70% as you said
but I know that it does not start with the 99.9%. We have been having
trouble getting reclaim to finish. Since it starts with the 70% and does
not finish, over time we have a lot of tapes offsite which are 99%
I think Wanda is correct except I though that if a management class does not
exist in the new domain then files in that management class will bind to the
default class in the new domain. This is what I remember but you can not
trust my memory. What was I talking about..
--
Phillip
320-4462
I can not speak for Solaris but for HP-UX and AIX, the last known state of
that file system is kept until remounted or manually deleted.
--
Phillip Ford
Senior Software Specialist
Enterprise Computing Services
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
Try an audit volume on it and see what happens. This usually takes care of
it. Sometimes you have to run the delete volume a couple of times.
--
Phillip
320-4462
-Original Message-
From: T. Lists [mailto:[EMAIL PROTECTED]
Sent: Thursday, July 28, 2005 1:58 PM
To:
If you are below that version we do it with a grip -v -e
dsmadmc directive | grep -v -e Highest -e Tivoli -e Command -e Copyright
-e Session -e Server
The above is from one of my scripts.
--
Phillip Ford
Senior Software Specialist
Enterprise Computing Services
Schering-Plough Corp.
(901)
Don't delete it. That will get rid of the copy as well. You need to do a
restore volume. Do a help on that. I usually do a preview and find out
what tapes I need from the vault first. I mark all the tapes as unavailable
so that TSM will not use them. When I get them all in the library I
Do you have a directory statement. I don't remember the exact wordage but
DSMC or something like that in the .sys file. If not then what you are
probably seeing is directories being rebound to the new class. If there is
not a DSMC command then directories bind to the class with the largest
Copytape reclamation generally uses copytapes if they are in the library and
not offsite. If they are offsite then reclaim uses the primary tapes. I
have seen your problem if the primary tape is marked unavailable for some
reason or is taken out of the library. It can also fail if the primary
Server commands must go to the server address ([EMAIL PROTECTED]) and
not the list address ([EMAIL PROTECTED])
I have attached instructions that were sent to me.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
I still use it also and would be lost without it. It is much easier to look
at things. I hate the web one.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From: Loon,
|
| Viaduktstrasse 42, P.O. Box, 4002 Basel - Switzerland |
| Phone: +41 61 226 19 78 / Fax: +41 61 226 17 01 |
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
-Original Message-
From: Ford, Phillip [mailto:[EMAIL PROTECTED]
Sent: Wednesday, 06 August, 2003 15:06
If there is no contents on the volume and it will not go back to scratch
then do the following:
Update vol 08 access=readw
Audit vol 08 fix=yes
Update vol 08 access=offsite
Now the volume should show as empty (or pending) and not filling. If you
are running DRM it will take care of
You say that there is another management class in the domain. Does it have
longer retention than the default. If so then the directory objects are
bound to this class and if this class is to tape then the directory objects
will go to tape.
Hope this helps
--
Phillip
320-4462
-Original
One way to get multi-restore to work is to collocate on filesystems in the
tape pool. If one collocates to the node then the problem that your
discussed happens. This is a designation of the pool so all nodes going to
this tape pool will be collocated by filesspace and that can eat a lot of
We saw this problem also. We ended up writing an OS script that the
operators start. It does the below command and then looks for one of two
things. If a request is found then it replies to it. If the move drm
process goes away then it stops. This has worked very good for us. But
like you
If you are doing it inside TSM I can give you an approach. I am not good
with SQL so you or someone on this list could help. You could run a
schedule everyday and have a select statement in it like the following:
select 1 from whatever where month(current date) = month(current date + 1
day)
We have TSM version 4.2.1.15 (moving soon to a 5... level) running on AIX
4.3.3. Another TSM manager told me that there should be a maximum of 16 DB
volumes. They said that there could be more but that they had heard that
there is system degradation if one was to use more that 16 DB volumes.
Here is my script
--snip---
#!/usr/bin/sh
#set -x
# file /var/operlib/adsm/remove_drm
# debug - 0 no debug 1-print 2-simulate
debug=0
lib=lib0
if [ $debug -gt 0 ]; then
echo debug is on and set to $debug
echo library to use is $lib
fi
cat /etc/adsm/adsm | read usr pw
Try audit vol 000592L1 fix=yes and then a delete if necessary. The answer
to this question is in the archives.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From:
We did not get a solution to this problem. We finally rebooted and ran
exchange optimize (perfwiz.exe) as requested by TSM support. This fixed us
for now. Just an update for everyone.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
I am going to stick my foot in my mouth but here goes. I do not feel that 3
months ago, 9 months ago, or 7 years ago are backups. These are archives.
We have the problem here that the higher ups tend to think of archives as
backups. We try to define backups for short term retrieval of data.
I have an open problem with TSM on this very subject. Their first incline
is to make sure service pack 4 is installed on exchange. On ours it is/was.
So that is not the fix. I will post if I find out anything. We feel that
it may be size related. We have too such clients. One small and
Here is a script that I use to remove media. It has to do a reply for each
tape. The section on looking for at q req and responding should do about
what you want. The functions at the top are from an old script and not used
in this one. I am short on time so I did not clean it up for you.
Backups
Creation of copypool volumes from disk
Creation of copypool volumes from primary tape
DB backup
Remove offsite tapes
Migration from disk to cartridge
Expiration
Reclamation
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901)
It uses both specs and it applies all of them. Thus, the answer to your
question is the 7th file will fall off after 30 days and there will only be
6 due to the following:
Retain Extra 30
Versions
rule. Also if you delete the file, the next day (or after a backup up and
expire
If you don't want to change the REUSEDELAY but force some tapes back to
scratch, I believe you can do a delete vol on them and they will got to
empty. If they are in the library, they will go to scratch. I use this on
my primary tapes when I run out of scratches and need a scratch tape for DB
Why not try
dsm -virtualnode=nodename
This works for us.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From: Schmitz Garnebode [mailto:[EMAIL PROTECTED]]
Sent:
Others can correct me if I am wrong. Extra versions (inactive) are still
controlled by the management class. The final active versions are kept
forever. TSM does not know if they were deleted or the filespace went bad
for some reason. Thus one has to hand delete them from TSM if they are not
I can tell you what we do. I make a script that is a place holder like
process_x_ok. In my main script at each point that I may want to stop I
check to see if there is a script call process_x_stop (see below). If yes,
then rename the script to process_x_ok and stop. If no then continue. This
Make the tape pool collocation enabled and do a migration on the disk pool
(update stg diskpool hi=0 low=0). Also helps to have migration process set
to 1 for the disk pool. I don't know if this is necessary or not.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
For this reason we run a backup stg from the primary disks to copytape pool
before we do migration. If the copytape is damaged by a tape drive, the
data is still on disk. If the primary tape is damaged during or after
migration then there is the good copytape.
So the question is do you have a
We just upgraded from 3.7.2 to 4.2.1.7 on our AIX 4.3.3 ML10 RS6000 TSM
server. We have started having some core dumps of TSM. About one to two a
week. I have found version 4.2.1.8 through 4.2.1.15 on the FTP site. Can
anyone tell me if any of these are stable? I remember someone telling me
Yes you can just do a cancel proc and it will die after it finishes the
current file. When you issue the backup stg command later, it will do
files that do not have a backup. We do this all the time when we have
service people come in.
Hope that helps
--
Phillip Ford
Senior Software
|
=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
-Original Message-
From: Ford, Phillip [mailto:[EMAIL PROTECTED]]
Sent: Friday, 16 August, 2002 16:41
To: [EMAIL PROTECTED]
Subject: Re: LTO Bulk Reader filled
When I tried the remove=untileefull clause (vers 3.7.2), it then asked me
to respond to each tape with a reply. If I have
I do it for the last Friday of the month. I do it by using the TZ variable
and number of hours for a week (168, 24 hours by 7 days). My script is
made to run on Friday only (every Friday). I get the current month(cm) and
I get next weeks Friday's month with the following for central time:
When I tried the remove=untileefull clause (vers 3.7.2), it then asked me
to respond to each tape with a reply. If I have to respond to each tape, I
might as well do it the old way and remove them when I count to ten. I want
it to take them out and not ask me and then stop when the bulk area is
We do it similar but add extra steps. We do the following:
1) in the stop script we do a nohup halt_script . This does the same as
you did below except we added the -noc switch on the dsmadmc line. I have
seen times when the dsmadmc line hangs and thus the script hangs. And then
shutdown can
We have only done this with our exchange servers (not the basic node but the
data from the exchange TDP). On these systems we have three nodes
base_node, base_nodeA, and base_nodeM. The base_node is for the operation
system. The A is the current exchange node. If told to lock, we will
change
I'm another that still uses the old GUI I feel for the ones that have
come after this GUI was dropped.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From:
Did the permissions change on the file or owner or group? If any of these
change, TSM will take a new copy of the file as if the file had changed.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
We use the following: this is the macro to update the actual macro:
--
delete script reclaim_check
def script reclaim_check desc='Check if space reclamation is finished'
upd script reclaim_check select * from processes where process='Space
Reclamation'
upd script reclaim_check
Try like instead of =.
select node_name, filespace_name from filespaces where node_name like 'AD1%'
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From: Neil Rosenberg
Try an audit vol 000324 fix=yes and then delete vol 000324 discard=yes
if it does not go to scratch on its own.
Hope that helps
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original
We sometime get this with a broke tape or a tape that is write locked. We
usually check them out and look at them. If they look good we log their
number and check them back in. We log the number to see if it happens to
them again. If so we get rid of them.
Just my 2 cents worth.
--
Phillip
You only need a reclaim pool if you have just one drive or the number of
mounts for the device class is set to one. If you have multiple drives and
define the device class to use drive or more than one then you do not need
to define a reclaim pool.
--
Phillip Ford
Senior Software Specialist
I don't know how to go back but when we went from single path to multipath
the ROM chips in the library had to be changed. So I assume that to go back
you have to get single path chips. The robot should be set to a different
lun but same target as a drive. What type of problem are you having?
For scsi I do the following before starting TSM:
tapeutil -f /dev/rmtxx reset
tapeutil -f /dev/rmtxx unmount /dev/null 21
This resets any scsi locks and unmounts any tapes.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901)
Here is a document in MSWord that I use.
Hope it helps.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From: Al'shaebani, Bassam [mailto:Bassam.Al'[EMAIL PROTECTED]]
After running your disk to copy backup ( backup stgpool primarypool
copypool), run a backup from onsite tape to copy tape (backup stgpool
onsitetapepool copypool). You should do this even if you do not have any
straight to tapes. What happens if a disk pool gets full and migrates or a
new file
We put on 18N2 and IBM said that there was a problem with it. They told us
to go to 22U0 which we did. We have had no problems with it as far as I
know. Don't know what type of problems you are looking for. We do not do
any straight to tape backups. All our backups go to disk pools and then
This is not true. Most of our UNIX clients are run from cron and we get the
session summary in the server actlog. We use this data to tell the clients
if they had any failed files. Our server is at 3.7.4 and we have various
clients from 3.1 to 4.2.
--
Phillip Ford
Senior Software Specialist
I live in Memphis. Our corporate office in New Jersey uses it extensively
for backing up their VMS systems. That is about all I know at their site.
I have tried to evaluate it here. I have mixed results. On the whole it
seemed to work very well and was easy to configure. I have two problems.
Looks go except need to do a database backup. We do it at 2.5.
All data in the offsite pool is also in the onsite pool. So TSM uses
whatever is available. If a copy tape is still onsite it can use it. If
not then the reclaim will use the onsite pool to create a new tape for the
offsite pool.
Try:
update admin passexp=days
If days are blank or not set (q admin) then I believe it is 30 days. If set
to 0 then do not expire password.
Hope this helps
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
This is not a flow chart but some things to do.
You can query cont to see if the tape has data on it. If it doesn't then
see if it is a DBbackup, look at volhistory.
If it does have data, then where is it? Query libv libx volume will show
if it is in a library. If yes, I then update vol xxx
Try
audit vol x fix=yes
and then
delete volume xx discard=yes
Hope that helps
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From: Alan Davenport
We have some very large oracle databases. Most of these compress very well.
We did test with and without client compression. The rule we used for best
is the total elapsed time. These were going to disk pools. We always got
better results with client compression than without. As an added
You did not say what type of system the client is. This was easy on our
unix boxes. The dsm.sys file is made to be in stanza format and thus we use
two different client nodes. One for the system and one for Oracle. Since
these are treated as two separate nodes to TSM, they do not get in each
We do this by declaring two separate nodes on the TRU64 systems. The base
node (base) is for the OS and all database files are excluded. We do backup
the redologs under base but use a different management class so that we do
not keep them as long. On the weekend we do our cold backups of just
Off the top of my head, have the first script submit another script that
does the checkout. Submit this script to run x minutes in the future. This
is a kluge but should work. I don't know of any way to delay in TSM
scripts.
Hope this helps
--
Phillip Ford
Senior Software Specialist
You have to set the password expire period to 0 days (that is zero) (use
update node passe=0 and update admin passe=0). The default is blank and
goes to 90 days or 60 I don't remember. I set this on the node and
administrative client for the node.
Hope this helps
--
Phillip Ford
Senior
We had this problem with disk cashing on. When the disk pool was full of
old data. Enough was cleaned for the estimated size. Then the client sent
a larger file (usually compressing a compressed file) and it would fail for
not enough space. We turned disk caching off and the problem never
I had a problem like this and it was because the management class name was
too long. I don't remember what version or *sm I ran into it with but I
shortened all my management classes to 10 character. I found this on a
class called performance which I shortened to performanc.
Hope this helps
We do this all the time. The script is called migrate_check as follows:
select * from processes where process=Migration
if(rc_ok) goto resch
delete schedule migrate_check t=a
expire inv
exit
resch:
delete schedule migrate_check t=a
define schedule migrate_check t=a cmd='run migrate_check' -
Find attached the steps that I use to fix broken/damaged tapes.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901) 320-4462
(901) 320-4856 FAX
[EMAIL PROTECTED]
-Original Message-
From: Cook, Dwight E [mailto:[EMAIL PROTECTED]]
Sent:
We have TSM 3.7.x running in an HACMP environment (AIX 4.3.3). I don't know
of any special setting we needed to set in TSM. TSM is not aware that it is
part of an HACMP. We made the TSM server directory on a disk array. This
lv is a resource that is switched between nodes as the TSM resource
Yes. This is how we have our three 3575s attached. You have to remove the
terminator from the pci cards and terminate on the end of the twintail. Our
RS6000 are H70 but I am sure that the H80 will be the same.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
You could logically divide up the client system in to two. Register two
nodes with *SM. Make two opt file sets (includes/excludes, pre and posts,
and proper stanzas in sys file) with proper domains and excludes/includes
(one for first half and one for the other). These should be able to run at
This is from memory but this has been asked before. As I remember ( and it
is going fast) you need to remove the spaces around the . Or maybe after
the .
Sorry that I don't remember more.
--
Phillip Ford
Senior Software Specialist
Corporate Computer Center
Schering-Plough Corp.
(901)
Sure we do it all the time. It goes
Disk-OffsiteDLT
Disk-OnsiteDLT
Now just to catch anything that has skipped or migrated from disk do
OnsiteDLT-OffsiteDLT
Just make sure you use the same copytapepool for all the commands.
This is how we do ours every day.
You can also do it by turning on
Hi Kelli - I use to live in Chesterfield County a long time ago (left in
1962), by the dam (Bolder Dam in I remember right) on the James river on
Cherokee road. I wish we had that place now.
In the include/exclude file you can tie a set of files/file to a management
class. Each management
The way we do it is to reschedule itself for ten minutes later. As follows:
select * from process where process='Backup Storage Pool'
if(rc_ok) goto resch
delete schedule this_sched t=a
now do my thing goes here
exit
resch:
delete schedule this_sched t=a
define schedule this_sched t=a cmd='run
From my knowledge the include/exclude on UNIX is read from the bottom to the
top. At the first match for a given file the search stops. Thus, in your
example, the bottom-most entry is in control and works for backup but not
during the archive (wrong management class). I would make separate
We have the same message on one HP11.0 system. This is what TSM said:
--
-Original Message-
From: Mark Shilling [mailto:[EMAIL PROTECTED]]
Sent: Friday, February 23, 2001 1:48 PM
To: Ford, Phillip
Subject: RE: PMR 19576,998
Yes we do this all the time.
First run backup from disk to copy pool
backup stg diskpool copytapepool maxpr=4
Then we do a backup of primary tape pool to same copy tape pool incase
overflow of disk pool happened:
backup stg tapepool copytapepool maxpr=4
Then finally we migrate the diskpool
This is what I use:
PS1=$(uname -a):${LOGNAME}':${PWD#$HOME/}:! '
export PS1
This looks like the following:
node_name:root:/tmp:300
The 300 on the end is the number in .sh_history that the next command will
get.
Hope this helps
--
Phillip Ford
Senior Software Specialist
Corporate Computer
Because of this (we are just looking into backupsets), we have decided to
use the description area to denote node name and date. Then this can be
used to tie the volumes together. I could not come up with anything else
yet. I will be interested in what the rest of the group has to say.
--
In our scripts os scripts we use the following:
cat /etc/adsm/adsm | read usr pw
then
dsmadmc -id=$usr -password=$pw q ses
the file /etc/adsm/adsm is protected from access for general users. Root
and one group that would know the password anyway are the only ones that can
read it.
That is
You wrote in the below command for the source
-fromnode=medrs2/ora01/apps//oracle/admin/ACWT/pfile/initACWT.ora
shouldn't be
-fromnode=medrs2/ora01/app/oracle/admin/ACWT/pfile/initACWT.ora
And the destination form command
/ora01/app/oracle/admin/ACWP/pfile/initACWP.ora
from below
I can not say what is best but what we do.
1) backup diskpool to DLTPOOL2
2) Migrate
3) backup the tape pool to DLTPOOL2
4) Backup DB
Both backups must be to the same copy tape pool.
We do it in this order to keep down tape to tape operations. If no backups
occur during step 1 or 2 then step 3
87 matches
Mail list logo
