)
-Original Message-
From: Steven Pemberton [mailto:[EMAIL PROTECTED]
Sent: Thursday, May 22, 2003 7:43 PM
To: ADSM: Dist Stor Manager; DFrance
Subject: Re: Question on defeating TSM strengths: due to budget
constraints solved
On Friday 23 May 2003 11:08, DFrance wrote:
With all due respect, I
Ed, Go check out OTG's Disk eXtender, now owned by Legato... it does the HSM for NT,
is not as spiffy as the native Unix version from IBM, but, that's what is out there,
at this time. Get the doc from their web site, after you register, you can get the
full doc. eval. software.
Don France
Ed,
You are right on both counts; backup objects have one set of retention parameters,
from the MC in the backup copygroup; backupset is one way to achieve different
retention (for all the active objects at the point-in-time it's created); another
method is to use more than one nodename,
Also, you might try running a backup using -dirsonly option? This, at least, would
re-establish the directory info -- you can then restore the data/file-tree structures,
but you'd likely have trouble using point-in-time parameters for a restore that
includes the lost (older) directory objects.
Some customers mitigate this security issue by eliminating the DSMCAD service, as a
matter of policy; that's probably okay for some businesses -- not likely okay for
help-desk when supporting desktop users.
A number of requirements are being considered (thru SHARE) along the lines of better
Regardless of what script/TDP/etc. is used, sending (two copies of)
archive logs to two separate tapes is done using TWO storage pools.
One way to do this is via copypool, which depends on timely execution of
an admin command; another way (most dba's prefer) is to use
-archmc=ARCH1 (and
Hi David,
Haven't seen this on v5.1.1.6 -- would sure like to know how it gets resolved... have
a customer running 5.1.6.2 since last week, and wants to start using server-prompted
mode (for more precise scheduling). Keep us posted, eh?!!
Thanks,
Don
Don France
Technical Architect -- Tivoli
Sadly, the events info (via q events) is a consolidation of several tables,
including internal info not externalized; I understand that IBM is working to resolve
this so we can select query such info.
For now, you must use the q ev command (with modifiers) to easily get this info; in
your
A) The speed of new tape drives (eg, 9840 3590) with their mid-point load mostly
mitigates the restore speed issue; even IBM's LTO or STK's 9940 seem to be
sufficiently fast, they're more like the speed of disk of just a few years ago;
B) TSM further mitigates restore speed by consolidating
Retention of 65 days means after 65 days -- so that would be tomorrow; but, even
then, only AFTER you run EXPIRATION will they be deleted from the db. To
prevent/defer that action, just avoid running EXPIRATION.
Unfortunately, there is not a supported (ie, externalized) mechanism to
To (briefly?) summarize... (the first two bullets address your original query):
- With TSM, all the active AND inactive versions are (a) always in the silo, (b) only
sent across the network ONCE (if you use the progressive-incremental the way it was
intended)... extra copies of the same version
This problem sounds impossible,,, kinda like an Ann Landers test(?).
If you want to restore file that changed AFTER the latest backup, by definition, it
has not been backed up, yet! I suspect, like other respondents, that you are really
looking for point-in-time restore...
I worked with a
The successful backups do NOT say all files were sent during backup; following
thoughts may help you discover the true reason for the missing files:
- C$ on WinNT (esp. Win2K) has lots of system-protected files that get skipped; they
are rolled into the system state blob, so could be all is
Sorry to say, you might have had a version with a big HOLE in it; both FILE and DISK
storage pool volumes must be local devices to Win2K... using 4.2.1.0. This may be a
problem with my 4.2.1 system -- as I recall, SANergy is designed to help this type
situation using SAN-based disk pools; it
You will probably need the nested form of the query, something like...
select a.volume_name from volumeusage where blah blah blah and -
a.volumn_name in (select b.volume_name from volumes where -
last_write_datecurrent date - 60 day)
We are requesting an
Hi Dave,
There seems to be some contradicting info -- DIRECTIO in Tech. Guide says
what you say applies to ASYNCIO; yet, in the Admin. Ref., it states no such
limitations. Then, for ASYNCIO, there's alot of verbiage about how to
configure, and no such stated limitations in either the Tech. Guide
The short answer is not yet...
I liked 4.2.2.10 (and .12); also, 5.1.1.6 is a good one.
Earlier today, someone pointed out, patch-level 5.1.5.3 was withdrawn, etc.
Don France
Technical Architect -- Tivoli Certified Consultant
Tivoli Storage Manager, WinNT/2K, AIX/Unix, OS/390
San Jose, Ca
(408)
Before proceeding, upgrade to 5.1.1.6 (or later); there were problems in
the 5.1 base, GUI-console (and web-admin) are not installed for proper
operation. Upgrade to 5.1.1.0, from ftp site, then add patch level that you
like (I would advise 6 or higher, imho).
HTH!
Don France
Technical
Gerald,
This is a known requirement, is being submitted to the developers thru
SHARE; a preview=yes option for restore/retrieve, which runs thru the
database as if the restore is being processed --- specifically, to obtain
the list of volumes that will be required.
So far, their response was
Another way to (possibly) cheat the system on this is to
(a) run the select from volumeusage to determine list of tapes,
(b) mark volumes outside the silo unavailable, then
(c) use classic restore --- get what you can *and* the unavailable vols!
Unless they closed this loophole, it used to be
Read the Unix client user's guide about -Archsymlinkasfile option... you
might (likely) want to change your use to the non-default behavior, No.
Don France
Technical Architect -- Tivoli Certified Consultant
Tivoli Storage Manager, WinNT/2K, AIX/Unix, OS/390
San Jose, Ca
(408) 257-3037
FYI, v5.1 supports server-2-server export-import in a single step; you no
longer need to run the intermediate data thru virtual volumes --- it works
abit like CMS-pipelines, once you get it setup on both sides, data flows
from server-a (node-data-on-tape input) to server-b
Yes... this is an excellent idea; it's so good, that IBM made it even
easier on AIX -- search for dsmulog in the admin. guide... it creates the
flat-file in real-time, can be setup to automatically roll-over to a new
file every night at midnight, so I usually advise 30-day retention on act.
log,
So... does this same constraint apply to the new Windows IMAGE backup?!?
In particular, I have a customer with Celera NAS,,, wondering if they could
do an image backup (offline or online) of a NAS share from external Win2K
box.
We already know NDMP is coming real soon, early next year -- not
-3037
mailto:don_france;ayett.net (change aye to a for replies)
Professional Association of Contract Employees
(P.A.C.E. -- www.pacepros.com)
-Original Message-
From: DFrance [mailto:DFrance-TSM;att.net]
Sent: Sunday, November 10, 2002 4:40 PM
To: 'ADSM: Dist Stor Manager'
Subject: RE
Nope... all the versions of the directories should be rebound (as indicated
by the messages in the dsmsched.log -- presume you use client scheduler to
run the daily backups).
BTW, did you verify that this helps you for your Netware environment (by
query the filespace to see if any data is getting
So, Allen -- what is it you do that accomplishes the (virtual) filespace
configuration... on NT/Win2K?
Don France
Technical Architect -- Tivoli Certified Consultant
Tivoli Storage Manager, WinNT/2K, AIX/Unix, OS/390
San Jose, Ca
(408) 257-3037
mailto:don_france;ayett.net (change aye to a for
So... thru most of this thread, the upshot is to consider:
1. Avoid cache=yes, rather use large disk pool with migdelay=1 *and* avoid
offsite reclamation -- maybe defer the offsite reclamation to weekends,
after clearing out the disk pools.
2. Use not one, but TWO tape technologies -- 3590 (J or
It wasn't a problem before (unicode client support); I've done exactly that
with WinNT, Win2K and Unix clients in prior versions. (In the old days the
vol-label on WinNT needed to be the same.)
As long as you didn't change the TSM client-level from non-unicode to
unicode, you should get what you
NT backup or TSM
2. Perform the authoritative restore by booting into DS restore mode and
running ntdsutil.
-Original Message-
From: DFrance [mailto:DFrance-TSM;ATT.NET]
Sent: October 30, 2002 1:18 AM
To: [EMAIL PROTECTED]
Subject: Re: More on Repeat Backups...
Actually, my personal
As of 9/22...
Yep... this is a problem -- which we have, also, recently experienced!
It appears to be related to a level-sensitive ssi (daemon) that runs on the
TSM server, arbitrating communication with the ACSLS server -- according to
Stephen, after discussing this with him (and researching
Check out the PASE announcement; I just attended a web-conf on the PASE for
OS/400 announcement, as it relates to TSM.
There's info on the BRMS web site and (only basic) info in the announcement
letter for v5.1 of TSM server (April 12, 2002).
Don France
Technical Architect -- Tivoli Certified
I like all what you two said; simply turn on the perform client trace
opt, specify output file, post the results --- it's a small file of output,
less than 50 lines --- then we can really discuss the issue. (BTW, perform
will include client_instr_detail along with time-stamps and client options
Andy,
Thanks for researching this; boy, I forgot how many versions got hit with
this... I might be abit foggy, myself, but I was supporting two accounts at
the time (Oct, 2000, again in April, 2001) -- one gottit (big time, several
TB) with the issue on 3.1.0.7 f2, the other was on 3.7.2 (but had
Nice response, Zlatko --- I was about to say some of the same!
I agree, and would like for Geoff to reply about what he sees, relative to
the GUI window (one clear indicator). Do his SAN drives appear under the
Local branch or Network? (I noticed at my old 4.2 level, my
locally-defined shares
To really know if you need a second server, you need to know what your
current server is doing. An M80 with 5GB of RAM and 2 drawers of SSA drives
should easily handle 1.5 to 2.5 TB/day; the wide variance is a function of
the number of db-inserts that occur. Do you know your current backup load?
So,,, the upshot is, for good performance of OFFSITE reclamation:
- empty the DISK (ie, RANDOM) storage pool(s) that feed any
offsite/copy-pool before running reclamation.
This, loosely translates to only run reclamation for offsite storage-pools
on weekends -- since we sized the DISK pools to be
From all the posts I've been reading, 4.2.anything is NOT where you want to
be -- the sunset date for 4.2 support is April 15, 2003 -- less than six
months away.
I am working with a customer who's on 4.2.2.12, we plan to upgrade to
5.1.1.6 -- which (currently) seems to be the cleanest 5.1
It was the client code that was previously fixed, more than once, in the 3.1
and 3.7 timeframe --- sounds like the DST code for Windows got broken,
again. 4.1.x is downlevel, for support contract purposes -- unless things
are different for you.
Don France
Technical Architect -- Tivoli Certified
Nope... q libv command will list all tapes in defined libraries; you can
wild-card the library name and/or the volume number. Seems to me that when
you do bulk-loading the library (ie, you open the door and load slots), then
run TSM audit libr, there's a missing step or 2 -- make sure no drives
Yep... most data center customers I've worked with prefer to do online full
backups during the week, then weekly and monthly offline/cold backups...
using dsmc archive /Ora_instanceX/* -archmc=xyz -su=y. Using archive
makes it very easy to do just what the customer wants: weekly (and/or
monthly)
When a restartable restore is currently running, it displays the session
number -1... only possible when NQR is invoked.
Once that number becomes positive (you have stopped the active restore),
then you can (from dsmc on the client) run restart restore (see the using
clients book). If you
ANR0486W Session session number for node node name (client platform)
terminated - internal error detected.
Explanation: The specified client session is ended by the server because an
internal processing error has been
detected on the server. A programming error may have occurred in the server
You must either (a) use TWO node names for this client (in order to have TWO
management policies), or (b) use the archive command (and -archmc=xxx) for
the carefully specified, long-term retention stuff, or (c) create a monthly
backupset for the filesystem in question.
Don France
Technical
Well... actually, if you really want to do it (I would consider it), you can
exclude System Object (or, better, use a special management class which
maps to a copygroup that has FREQuency=5, so INCRemental backups only occur
every 5 days).
Clearly, the 200-400MB blob called System Objects,
Tim,
Just use your keys from 5.1.0.0 media -- see the 5150 or 5151 readme files;
the only folks needing new media (at 515 level) would be for the new Linux
server, its material is in rpm format.
For all other servers, 5.1.5.0 and 5.1.5.1 are both available at the IBM ftp
site. BTW, if you
@IBMUS
Internet e-mail: [EMAIL PROTECTED] (change eye to i to reply)
The only dumb question is the one that goes unasked.
The command line is your friend.
Good enough is the enemy of excellence.
DFrance [EMAIL PROTECTED]
Sent by: ADSM: Dist Stor Manager [EMAIL PROTECTED]
10/26/2002 14:27
Please
Geoff,
Yep... there have been several other reports about 5.1.5.1 -- looks like the
oven was still not hot enough for this one... it's still not quite fully
cooked. We knew base 5.1.5.0 was flawed from IBM bulletins; didn't know
the full extent of the issue, though.
Last I checked, 5.1.1.6 was
I'd heard that 5.1.5 vanilla release was flawed (big surprise!?!) Maybe you
can try the first PTF (5.1.5.1), or patch level, for your situation...
Version 5 Release 1, Level 5.2 PTF
IP22546_02
(this one just released... 10/17)
There's been a long history
Tapes marked unavailable usually just need to be marked available (and
possibly checked back into the library). I suggest setting it to readonly,
for now.
If the msg you got is waiting for files from the server... that just
indicates no-query-restore is invoked, the client will be receiving all
Alex,
If the tape is truly bad, first audit that tape (audit vol xxx fix=yes);
if you just want to skip past that tape, one trick I've used is to force
classic restore -- nqr won't skip unavailable tapes, classic will (or it
used to in the v 3.1 timeframe)...
To force classic restore, specify
Halleluiah!!! I have ALL these same complaints, too; I hope our friend in
Virginia sees and responds to this! It's a great service he does, but it
needs to be made usable, again.,,, nice experiment, now let's fix it,
please?!?
Don France
Technical Architect -- Tivoli Certified Consultant
Fantastic... this is excellent resolution -- the old screen is still there,
just gotta know its new address!
Thanks,
Don
Don France
Technical Architect -- Tivoli Certified Consultant
Tivoli Storage Manager, WinNT/2K, AIX/Unix, OS/390
San Jose, Ca
(408) 257-3037
mailto:don_france;att.net
I share Mark's sentiments... on TWO points:
- if this was resolved, for both users in the discussion, let's share the
results;
- if DIRMC is no longer relevant, I'd sure like to know why! Win2K is
becoming ever more prevalent, and most data center customers go wild with
lots of permission groups
54 matches
Mail list logo
