Cannot comment on the drive "view" or schedule-mode problems -- sounds like
a TSM-cluster service definition discrepancy (else, a bug; need to check the
cluster services setups and client code level, then contact Support Line or
install latest client code -- there's been a bunch of client-code activity
in cluster support this year, both for Win2K and AIX).
Regarding your speed/performance of incremental {If the EMC disks appear as
"local" drives, you should consider the NTFS journaling-incremental
feature.} -- alternatively, you may want to consider -incrbydate for your
weekday backups; the speed of progressive-incremental is largely due to the
client traversing the entire file system structure to identify which files
to process -- file systems with large numbers of files (anything over
half-million) seem to be cause for performance concerns. I had a client
that decided to address this issue by limiting their file systems to 100 GB;
starting a new drive-letter when reaching that size greatly helped mitigate
the daily incrementals (AND full file system restores, their main concern).
We recently did 1.6 million file/object restore for a 320 GB file system,
achieved nearly 10 GB/Hr with parallel restore sessions, DIRMC (very
important), DIRSonly and FILESonly options, to minimize NTFS re-org
thrashing, and CLASSIC (vs. no-query) restore path, to ensure minimal tape
mounts.
Result was backups of 12-20 GB, restores 6-15 GB/Hr, depending on
collocation standards. High priority (ie, mission critical or
high-visibility) servers get collocated.
Hope this helps. See, also, a dozen other posts this past couple months,
search for "cluster" in the subject field.
Don France
Technical Architect - Tivoli Certified Consultant
Professional Association of Contract Employees (P.A.C.E.)
San Jose, CA
(408) 257-3037
[EMAIL PROTECTED]
----- Original Message -----
From: "Firl Debra K" <[EMAIL PROTECTED]>
To: <[EMAIL PROTECTED]>
Sent: Thursday, April 04, 2002 2:07 PM
Subject: TSM on a win2k cluster
> How is everyone elses experience using TSM on a win 2000 cluster?
>
> Here is ours.
>
>
> We have 2 win2k clusters that are just used for file/print. All the disks
> are EMC DASD connected via fibre. One (clusterA has 11 group-disk
> resources, the other 9 (clusterB). The clusters are used in any mode,
> active-active any drive combination, active-passive.
>
> ClusterA (308g used of 572g), about 4 million files
> ClusterB (642g used of 719g), about 4 million files.
>
> Both clusters are just used by users to shares for file and print access.
>
> Are there others out there that have clusters of this capacity and using
TSM
> to backup them?
>
>
> First Attempt.
> I looked at the redbook, sg24-5742-00, Using Tivoli Storage Management in
a
> clustered Windows NT environment.
> I first configured it using the common names method. Initial complete
> backups of volumes took a while. We averaged 4-6g/ hour.
> Worked somewhat ok, tell I tried to kick off the scheduler. When the
> cluster is in active passive mode, I could get the scheduler to work using
> sched mode prompted ,and refer to one dsm.opt file that included all the
> cluster domains, e:-o:., It would grab the last scheduler that was started
> and look at the dsm.opt file and run from there. So one schedule ran to
> include all the drives.... NOW when the cluster is in active\passive mode,
I
> could only get it to work with sched polling and not all the schedulers
> would kick off. 2 to maybe 4 would kick off of the 11, so when using a
> separate dsm.opt file per disk cluster group only.
> Dealt with tech support, never got more scheduler to kick off then the 4.
> NOTICE that I had to have different configurations depending on how the
> cluster was in, whether active-active or active-passive. So that was not
a
> solution.
>
> Tech support suggested using the unique names method which is the only
> method now referenced in their newer documentation.
>
> Ok, I registered with TSM a separate node per cluster group. Created a
> separate scheduler per cluster group. So I have 11 plus the quorum. The
> initial complete backup speed about 4/6 g... .. Now something interesting
is
> happening. One one node 2 of the cluster group disk can't be seen in TSM.
> The users are using it fine on the server and it is available in the
> operating system. I tried enabling the scheduler for those 2 disks, TSM
> does not back them up because it does not see them. I swap to the other
> node the disk cluster groups. TSM can now see it and manual backups and
> scheduled backups works. Weird!
> On the node that does not recognize the disk, thinks they are not
clustered
> disks for some reason. Noticed the error when doing a command line backup.
>
> The questions.
>
> Do other business have clusters like these with this amount of disk space
> and successfully using TSM to do backups? Are others experiencing these
> type of issues?
>
>
> Issues we see is the performance of the incrementals using the common
names
> method is dreadfully slow. Sometimes take days to finish a incremental
even
> though their may not be that many changed files.
>
>
> Incrementals using the unique method is faster, since each one has its own
> dsm.opt file and there are separate services that are able to run.
>
>
> Btw-we do use fdrsos to backup the EMC disks, but it is using tpo many
tapes
> and getting quite costly. That is why we are trying to use TSM.
>
>
> Just collecting information on how successful other groups are and what
type
> of setups their clusters are in.
>
>
> Thanks in advance.
>
> Debra Firl
> Mellon Financial Corp.
>
>
>
>
>
> *****************************************************************
> DISCLAIMER: The information contained in this e-mail may be confidential
> and is intended solely for the use of the named addressee. Access,
copying
> or re-use of the e-mail or any information contained therein by any other
> person is not authorized. If you are not the intended recipient please
> notify us immediately by returning the e-mail to the originator.
