backup performance with db and log on a SAN
I recently moved the 36G TSM database and 10G log from attached SCSI disk drives to a SAN. Backing the db now takes twice as long as it used to (from 40 minutes to 90 minutes). The old attached disk drives are non-RAID and TSM mirrored. The SAN drives are RAID-5 and TSM mirrored. I know I have to pay a penalty for writing to RAID-5. But considering the massive cache of the SAN it should not be too bad. In fact, performance of client backups hasn't suffered. However, the day after the move, I noticed that backup db ran for twice as long. It just doesn't make sense it will take a 100% performance hit from reading from RAID-5 disks. Our performance guys looked at the sar data and didn't find any bottlenecks, no excessive iowait, paging, etc. The solution is to move the db and log back to where they were. But now management says: "We purchased this very expensive 2T IBM SAN and you are saying that you can't use it." Meanwhile, our Oracle people happily report that they are seeing the performance of their applications enjoy a 10% increase. Has anyone put their db and log on a SAN and what is your experience? I have called it in to Tivoli support but has yet to get a callback. Has anyone noticed that support is now very non-responsive? server; AIX 4.3.3, TSM 4.2.1.15 Thanks, Eliza Lau Virginia Tech Computing Center 1700 Pratt Drive Blacksburg, VA 24060 [EMAIL PROTECTED]
test for DRM
Hi there and thanks in advance for all your tips and recommendations we have a huge distributed new TSM setup, with server spread across the campuses. We recently moved from 3 ADSM 3.1 servers to 9 TSM 4.2.2 servers all direct attached SUN 280r(sol-2.8), SUN T3 and Spectralogic 64K libs I have a few questions 1. We have TSM working on Solaris2.8 with SUN T3 storage for mirrored DB and storage pools. Our performances is nowhere close to what SUNs recomended T3 sustained writes which is 80MB. Recovery logs are on external D130 disk-packs has anyone seen a setup with SUN and is this normal? My writes to diskpools are at 20 to 30 MB and that is slow. I have a raid5 setup for the storage pools, Tivoli suggests JBOD for storagepools rather than raid5!? but how do I protect myself from a disk fail on a critacal quarter financial backup.. since the source gets overwritten as soon as they throw the data on to my stoarge pools T3 (primary) 2. One such setup has a StorageTek L7000 lib and my customer wanted me to prove that the tapes from offsite do work. Tivoli suggests that I do not test DRM on a production system. But I had no choice but to atleast test for bad media on the primary tapepool!if any so I went ahead picked *a* node with a select statement, marked all the tapes destroyed on the primary tape pool(for that node), and started a restore of a filesystem. Prior I had a bunch of tapes recalled from the off-site pertinent to the same node. Had them checked in as private and waited, to see if TSM picks those tapes since the onsite were marked destroyed. This process has been rather lengthy and tedious and unsuccessful Has anyone done a rather simpler test for bad media, to prove that the off-site tapes do work, less to say the test I performed came back with data integrity errors and my customers are not happy and with all traces setup.. Tivoli was unclear how that happened (Tivoli claimed, there could be a flaw in my DRM process) 3. The last question, during a copy storage pools process, if I *cancel* the process (since it took days), the next time I start (manual or via a script) does it pick up from where it stoped! thanks for all your responses, forgive me, My knowledge is pretty limited and I started learning Tivoli while I started this project Cheers.. Chetan
Re: Hoping to get new adsm server
We did a simple cpu/disk upgrade during first shift. The libary was the same, and we moved from nt4 running on a Dell 6100 (dual 200Mhz) to a Dell 2400 also a dual cpu running w2k. After the last of the night's backup finished we copyed all the primary data to the offsite pool. Migrated all data from disk storage pools to tape, This saves an audit. Made a full db copy and an incr as some of the prior work overlapped. Copied the volhist, devcnfg and several other historic files to the new box.. Renamed the nt4 server to a new nodename. Move scsi-interface cards from the old server to the new server. Renamed the pre-staged w2k box to the old adsm/tsm server name as this was easl ier then changing dns in our case. Restored the data base tapes. Fixed up the storage pools by deleting the old volumes and creating new ones. Fixed up the libary by deleting the drives and libray and defining them again. I tried to use the new mmc wizard but the libary and drive names did not match those defined on the old system. So I used the old line mode commands from the orginal vm product days. The pc hardware and os changed, but the tsm software level and tape hardware did not. In the future, I would delete the disk storage pool volumes and tape devices before the move. Moving to a server lets you try a small subset of the skills required for a dr drill. len
Re: test for DRM
What SUN probably did not tell you was they were probably 32K or 64K blocks to get that data rate. TSM Database uses 4K and most other databases are 4K or 8K. Are you using raw volumes for the disk pools? If so, talk to SUN to find out what the optimal block sizes are. If you are using a file system, same thing. Be careful how you talk to Tivoli about RAID-5 versus JBOD. They think of RAID-5 as being a bunch of disks you have raided, not a hardware raid solution with a high end controller cache. 20MB to 30MB/sec sounds about right for a T3. The T3 is a midrange device. It may not be a good choice for a high write activity workload versus a JBOD/Raid-1. I do not know how smart the T3 is. The 280 is a pretty good little box based on my review of it, so I do not know that is the problem. You are preaching to the choir on a protected disk pool. If you lose it you have lost the backups for that night which may be unacceptable. One thing to consider is large files you may want to send directly to tape. You can do this by setting a maximum file size in the primary disk pool. TSM will mount up a tape and any time a file from a client exceeds the limit, it writes it to tape instead of disk. But, unless you have very reliable tape and create your backup storage pool copies in time to recapture the backup if the tape is bad, I do not know if this is an option for you. Backup Storage Pool commands start where they left off. You say your environment is huge. How many tapes do you have. How much are you trying to backup to these servers. They may not be the right fit. As far as the data integrity issue. Did you not back something up? Where you missing some data? What do you mean? This is one place TSM really shines over the other backup products. Paul D. Seay, Jr. Technical Specialist Naptheon Inc. 757-688-8180 -Original Message- From: Chetan H. Ravnikar [mailto:[EMAIL PROTECTED]] Sent: Saturday, August 31, 2002 12:09 PM To: [EMAIL PROTECTED] Subject: test for DRM Hi there and thanks in advance for all your tips and recommendations we have a huge distributed new TSM setup, with server spread across the campuses. We recently moved from 3 ADSM 3.1 servers to 9 TSM 4.2.2 servers all direct attached SUN 280r(sol-2.8), SUN T3 and Spectralogic 64K libs I have a few questions 1. We have TSM working on Solaris2.8 with SUN T3 storage for mirrored DB and storage pools. Our performances is nowhere close to what SUNs recomended T3 sustained writes which is 80MB. Recovery logs are on external D130 disk-packs has anyone seen a setup with SUN and is this normal? My writes to diskpools are at 20 to 30 MB and that is slow. I have a raid5 setup for the storage pools, Tivoli suggests JBOD for storagepools rather than raid5!? but how do I protect myself from a disk fail on a critacal quarter financial backup.. since the source gets overwritten as soon as they throw the data on to my stoarge pools T3 (primary) 2. One such setup has a StorageTek L7000 lib and my customer wanted me to prove that the tapes from offsite do work. Tivoli suggests that I do not test DRM on a production system. But I had no choice but to atleast test for bad media on the primary tapepool!if any so I went ahead picked *a* node with a select statement, marked all the tapes destroyed on the primary tape pool(for that node), and started a restore of a filesystem. Prior I had a bunch of tapes recalled from the off-site pertinent to the same node. Had them checked in as private and waited, to see if TSM picks those tapes since the onsite were marked destroyed. This process has been rather lengthy and tedious and unsuccessful Has anyone done a rather simpler test for bad media, to prove that the off-site tapes do work, less to say the test I performed came back with data integrity errors and my customers are not happy and with all traces setup.. Tivoli was unclear how that happened (Tivoli claimed, there could be a flaw in my DRM process) 3. The last question, during a copy storage pools process, if I *cancel* the process (since it took days), the next time I start (manual or via a script) does it pick up from where it stoped! thanks for all your responses, forgive me, My knowledge is pretty limited and I started learning Tivoli while I started this project Cheers.. Chetan
Re: backup performance with db and log on a SAN
You say you put it on SAN. What kind? I am guessing a Shark or FastT device since it is an IBM SAN. If you only got 10% with Oracle and the issues with TSM on the database I think you have your file systems on the AIX machine setup incorrectly or another problem. It could be the tape drive is the problem on the backup, especially if you are using the same HBA for both a SAN Data Gateway (IBM 3583) or other similar solution and the SAN disk and running client backups at the same time. Paul D. Seay, Jr. Technical Specialist Naptheon Inc. 757-688-8180 -Original Message- From: Eliza Lau [mailto:[EMAIL PROTECTED]] Sent: Saturday, August 31, 2002 11:19 AM To: [EMAIL PROTECTED] Subject: backup performance with db and log on a SAN I recently moved the 36G TSM database and 10G log from attached SCSI disk drives to a SAN. Backing the db now takes twice as long as it used to (from 40 minutes to 90 minutes). The old attached disk drives are non-RAID and TSM mirrored. The SAN drives are RAID-5 and TSM mirrored. I know I have to pay a penalty for writing to RAID-5. But considering the massive cache of the SAN it should not be too bad. In fact, performance of client backups hasn't suffered. However, the day after the move, I noticed that backup db ran for twice as long. It just doesn't make sense it will take a 100% performance hit from reading from RAID-5 disks. Our performance guys looked at the sar data and didn't find any bottlenecks, no excessive iowait, paging, etc. The solution is to move the db and log back to where they were. But now management says: "We purchased this very expensive 2T IBM SAN and you are saying that you can't use it." Meanwhile, our Oracle people happily report that they are seeing the performance of their applications enjoy a 10% increase. Has anyone put their db and log on a SAN and what is your experience? I have called it in to Tivoli support but has yet to get a callback. Has anyone noticed that support is now very non-responsive? server; AIX 4.3.3, TSM 4.2.1.15 Thanks, Eliza Lau Virginia Tech Computing Center 1700 Pratt Drive Blacksburg, VA 24060 [EMAIL PROTECTED]
Re: How to list files backed up
>From the server, this may be useful: select filespace_name, hl_name, ll_name from backups where node_name='YOUR-NODE' and type=file and state=active_version and backup_date >current_timestamp-1 day This one actually runs pretty fast because the node_name, type, and state columns are indexed. Paul D. Seay, Jr. Technical Specialist Naptheon Inc. 757-688-8180 -Original Message- From: Stewart, Curtis [mailto:[EMAIL PROTECTED]] Sent: Friday, August 30, 2002 9:51 AM To: [EMAIL PROTECTED] Subject: Re: How to list files backed up Look in the log files. On my servers it's dsmsched.log. Here's a snippet of one of mine. 07/30/2002 20:09:09 Normal File--> 6,956 USERS:/USERS/username/2017A/2017a-1/5c913c10._gr [Sent] 07/30/2002 20:09:09 Normal File--> 529 USERS:/USERS/username/2017A/2017a-1/5c913c10._hl [Sent] 07/30/2002 20:09:09 Normal File-->21,766 USERS:/USERS/username/2017A/2017a-1/5c913c10._is [Sent] 07/30/2002 20:09:09 Normal File--> 2,878 USERS:/USERS/username/2017A/2017a-1/5c913c10._pf [Sent] 07/30/2002 20:09:09 Normal File--> 4,621 USERS:/USERS/username/2017A/2017a-1/63f52efb._df [Sent] 07/30/2002 20:09:09 Normal File--> 6,226 USERS:/USERS/username/2017A/2017a-1/63f52efb._gr [Sent] Cheers curtis stewart -Original Message- From: Cahill, Ricky [mailto:[EMAIL PROTECTED]] Sent: Friday, August 30, 2002 2:53 AM To: [EMAIL PROTECTED] Subject: How to list files backed up I must be missed something really obvious as all I want to do is list all the files that were backed up by a node in it's last backup, but can't seem to find any simple way to do this. Heelp Thanks in advance ..Rikk Equitas Limited, 33 St Mary Axe, London EC3A 8LL, UK NOTICE: This message is intended only for use by the named addressee and may contain privileged and/or confidential information. If you are not the named addressee you should not disseminate, copy or take any action in reliance on it. If you have received this message in error please notify [EMAIL PROTECTED] and delete the message and any attachments accompanying it immediately. Equitas reserve the right to monitor and/or record emails, (including the contents thereof) sent and received via its network for any lawful business purpose to the extent permitted by applicable law Registered in England: Registered no. 3173352 Registered address above
Re: Moving Stgpools question
Actually, there are more restrictions than what Mathew indicates. A copy pool tape must be moved to the same copy pool. It cannot move the data to a new copy pool like a primary can move to a new primary. To move all your Copy Pool data to another copy pool, just build a new copy pool using all the primary pools and the backup stg command, then delete the volumes discarddata=yes in the original. The other piece is if you do not want to change your reuse delay you can always do a delete volume command to return a tape to the scratch pool. Sorry, for the incomplete explanation the first time around. You do need to do a help on MOVE DATA and understand it completely. Paul D. Seay, Jr. Technical Specialist Naptheon Inc. 757-688-8180 -Original Message- From: Large, Matthew [mailto:[EMAIL PROTECTED]] Sent: Friday, August 30, 2002 4:50 AM To: [EMAIL PROTECTED] Subject: Re: Moving Stgpools question Correction: Move data works on both primary and copy storage pools, but it will only move the data to a pool in the same position in the storage hierarchy. ie primary pool to primary pool, copy pool to copy pool. just my fraction of a euro.. Matthew Large TSM Infrastructure Engineer Lavington Street Int: 7430 4995 Ext: +44 207 902 4995 -Original Message- From: Seay, Paul [mailto:[EMAIL PROTECTED]] Sent: 29 August 2002 18:39 To: [EMAIL PROTECTED] Subject: Re: Moving Stgpools question Select 'move data ', volume_name, ' reconstruct=yes', ' stg=newstgpool' from volumes where stgpool_name='OLDSTGPOOL' > /temp/macrofile Notice the where clause is in caps. This will create a file of output that you will have to delete the first couple of lines from the output file as shown below: Unnamed[1] VOLUME_NAMEUnnamed[3] Unnamed[4] -- -- --- move data 421999 reconstruct=yes stg=newstgpool Next you will do a "macro /temp/macrofile" command. MOVE DATA only works on primary storage pools to a different primary storage pool. And, every tape has to be mounted and the data copied to the new pool. I would set your reuse delay on the source pool to 0 so the tapes will immediately scratch after the moves complete. Paul D. Seay, Jr. Technical Specialist Naptheon Inc. 757-688-8180 -Original Message- From: David Gratton [mailto:[EMAIL PROTECTED]] Sent: Thursday, August 29, 2002 12:50 PM To: [EMAIL PROTECTED] Subject: Moving Stgpools question I have several storage pools in my environment and I would like to consolidate them all into one storage pool (same dev type). My question is , is there any easy way to do this or do I have to do a "move data $volume stg=newstgpool" for each volume of every storage pool? I guess I could write a script to do it but I was hoping there was a was to move a full storage pool with one commandAny help would be appreciated.. Dave Gratton IBM Global Services --- This e-mail is intended only for the above addressee. It may contain privileged information. If you are not the addressee you must not copy, distribute, disclose or use any of the information in it. If you have received it in error please delete it and immediately notify the sender. evolvebank.com is a division of Lloyds TSB Bank plc. Lloyds TSB Bank plc, 71 Lombard Street, London EC3P 3BS. Registered in England, number 2065. Telephone No: 020 7626 1500 Lloyds TSB Scotland plc, Henry Duncan House, 120 George Street, Edinburgh EH2 4LH. Registered in Scotland, number 95237. Telephone No: 0131 225 4555 Lloyds TSB Bank plc and Lloyds TSB Scotland plc are regulated by the Financial Services Authority and represent only the Scottish Widows and Lloyds TSB Marketing Group for life assurance, pensions and investment business. Signatories to the Banking Codes. ---
Re: Help topics are strung across page
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of Sam Haziprodromu >Hi 'TSM ers, we recently upgraded our TSM server from 4.1 to 4.2 but we ran >into a problem. AT 4.1 when we sought help it gave us a full page and >syntax. On the newer 4.2 version help is now a single horizontal >line strung across the page. Could you please help us. You didn't mention what version of OS your TSM server is on. If you're on AIX, the word from Tivoli support is that you need to force a reinstall of tivoli.tsm.msg.en_US.server -- Mark Stapleton ([EMAIL PROTECTED]) Certified TSM consultant Certified AIX system engineer MCSE
Re: tsm client is down-level with this server version
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of Roger Deschner > We've got half a dozen clients stuck like this right now. The first case > I had was an important, tenured, and extremely impatient professor who > had converted from Win 98 to Win XP, decided that XP "stinks", and > wanted to format his hard drive and restore his comfortable old Win 98 > system from ITSM. (This is what backup is for, right?) Because this was > basically a point-in-time restore, deleting the node was not a possible > strategy. He was incredulous that it took me several days of > communication with IBM support to straighten it out and peppered me with > emails demanding that I work faster on it the whole time I was > exchanging special commands and their outputs. Sounds to me like you need an SLA that stipulates what happens when people backgrade *anything*. Something in the realm of "expect problems" and "no time limit on fixing things"... > The second was an aggressively confused user who was backing up two > computers using one node, and who thereby un-did the fix mere hours > after I had spent several hours with IBM Support fixing it. I refused to > fix it again for this user and made their node restore-only until I > communicated with their supervisor about our one computer per node > policies. *sigh* ...and another patch on the SLA that voids your responsibility when the user does something that none too bright. > I really wish the existence of a correcting APAR was made more public. > I'm going to install it soon (today!), but I'll still have those half > dozen clients who appear to have "unicode cooties" to untangle. > > Would it be possible to provide a safe and tested script, or an APAR > that introduces a new secret command, that can untangle this mess in > some automated way? The current method of fixing it is way too costly. If I had a script that I was told would either fix everything or destroy the database, I'd have to put that puppy back in its box and say, "No, thank you." Every TSM environment is different (often radically so). It's a very complex piece of software, and, being almost infinitely configurable, is also almost infinitely capable of destruction. You're a better man than I, Gunga Din. I wouldn't work in an academic shop for any amount of money. -- Mark Stapleton ([EMAIL PROTECTED]) Certified TSM consultant Certified AIX system engineer MCSE
Re: tsm client is down-level with this server version
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of Zlatko Krastev/ACIT > Looking at the length of this thread and shaking my memory this question > might be a good candidate for the Monthly FAQ (Mark, what do you think ?) Great minds think alike! ;o) -- Mark Stapleton ([EMAIL PROTECTED]) Certified TSM consultant Certified AIX system engineer MCSE
Re: Hoping to get new adsm server
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of Roger Deschner > Note we selected SuperDLT over LTO or other tape technologies solely > because SDLT drives could read our old DLT-IV tape cartridges. An SDLT > drive, however, cannot write to a DLT-IV cartridge as I understand it. You could have gotten around that fairly easily. Quick outline for migration from one server/library combo to another: 1. Set up new server/library. 2. Bring new server/library online. 3. Use old server/library for restores only, while new library/server does backups. 4. Allow old server/library data to expire away. Eventually, you'll have to move data from old to new, but you can do that once the amount of data in the old system has been whittled away to a manageable level. -- Mark Stapleton ([EMAIL PROTECTED]) Certified TSM consultant Certified AIX system engineer MCSE
Re: Exchange bases ...
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of Arni Eggertsson > Apparently Tivoli decided to use the Microsoft API to backup the > Exchange database, this is done to be able to do a "consistant" > backup of the database. It goes deeper than that. If you stick to the APIs within an OS and the applications that run under that OS, you never get bitten by the "we added a service pack, and now the backups aren't happening" bug that ArcServe and Netbackup suffer from constantly. Also, in the case of Exchange, Microsoft will not support an Exchange environment that has experienced mailbox-level backups and restores, since the Exchange APIs don't support them. -- Mark Stapleton ([EMAIL PROTECTED]) Certified TSM consultant Certified AIX system engineer MCSE
Re: Restore Performance.
From: ADSM: Dist Stor Manager [mailto:[EMAIL PROTECTED]]On Behalf Of Gallerson,Charles,GLENDALE,Information Technology > Does TSM 5.1.0.0 provide MOVE NODEDATA and multithreaded restores Yes. Read the 5.1 Administrator's Guide. It is available on the 'Net. (See my FAQ posted on August 24, question 03-02, to get the address.) -- Mark Stapleton ([EMAIL PROTECTED]) Certified TSM consultant Certified AIX system engineer MCSE
