Bye for now
Just a note to tell those few who know me that I'll be unsubscribing shortly (and to those who don't, apologies for wasting bandwidth). I expect to be doing more TSM work sometime next year so I may be back, but in the meantime... well, I know where the archives are if I need them. So long, and thanks for all the fish. And the help. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] Wars are won before they are fought - through preparation, attitude, strategy and the selection of proper allies. Delenn, Babylon 5, by J Michael Straczynksi
Re: What if?....
Sam Schrage asks: What is the feasibility of making a copy of my offsite tapes via the AIX command 'tcopy', or other utility, rather than using TSM--another storage pool, doing the backup, etc.--and take this 'off-site' set of tapes to a diaster recovery test? Any suggestions? Are you using barcode labelling? If so you would presumably need duplicate barcodes. You would need to be very careful about tracking the copies and the originals, otherwise you could be letting yourself in for some interesting and fun media management issues. Also, you would want to make sure you overwrite the TSM labels on the extra copies afterwards. Apart from those things, I can't see why it wouldn't work. As long as the utility copies the whole tape including the TSM label. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: Raw Devices
With raw you also have the risk that someone unfamiliar with the box will come along and think it's spare disk, and use it to grow a filesystem. That's the main reason I don't use raw disk. -Original Message- From: Cook, Dwight E [mailto:[EMAIL PROTECTED]] Sent: Thursday, August 16, 2001 6:46 AM To: [EMAIL PROTECTED] Subject: Re: Raw Devices I have both in use... with JFS file systems you get the benefit of journaling... with raw you get the benefit of having the entire device available for use reduction of the overhead of journaling and with raw you don't have to spend time doing dsmfmt against them... Dwight -Original Message- From: Eduardo Martinez [mailto:[EMAIL PROTECTED]] Sent: Wednesday, August 15, 2001 1:23 PM To: [EMAIL PROTECTED] Subject: Raw Devices Hi *SMers. Anyone of you has worked with raw devices on UNIX instead of defining your storage areas with dsmfmt over filesystems? Do I have any improvement by using them or do I have to avoid their use at all? Thanks in advance. Regards. = Do or Do Not, there is no try -Yoda. The Empire Strikes Back
Re: Query report
Mike Anderson wrote: I have a group that wants the following information for their backups: Client name, the start time of the backup, end time of the backup, and any files that missed. I tried to show them the schedule log, but this was not clean enough for them. Anybody have any ideas? The TSM server is AIX 4.3.3 the client nodes are NT 4.0 I see two options, depending on whether you want to produce this report on the client or on the server. 1. Use Perl or perhaps VB to process the schedlog and produce what they want. This would probably be done on the client, as a scheduled job. Information about Perl for NT can be found at http://www.cpan.org/ports/index.html#win32 2. Use an SQL query on the summary table. This doesn't give the names of failed files, but it does say how many, which could then trigger a business process to investigate. The query would be something like: select entity as node, - substr(cast((start_time) as varchar(26)),12,8) as started, - substr(cast((end_time) as varchar(26)),12,8) as ended, - failed as files_failed - from summary - where substr(cast((start_time) as varchar(26)),1,10) - = substr(cast((current_timestamp) as varchar(26)),1,10) - and activity='BACKUP' This gives you information about today's backups. It works in my situation because backups start after midnight. Your mileage may vary. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
lbtest utility - reference?
Is there a reference somewhere for the lbtest utility? I've looked in the obvious places (Tivoli doco, Richard's quick facts) and can't find any detailed info. The ACSLS query commands seem fairly self-explanatory, but what I'm wondering is whether the other commands are relevant, and if so how to use them. TSM 3.7.4 with STK Powderhorn library. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Auditing a shared library
Hi all, Quick question - is it safe to run audit library if I'm sharing the library with another app (Netbackup), and it's a StorageTek ACSLS-controlled library WITHOUT user access control enabled? I need to do this because I know a tape has been physically removed without being checked out. I have no physical access to the library so I can't just put it back in (and besides, it may be damaged, as it was jammed in a drive). I REALLY don't want to find all the Netbackup volumes in my inventory. It would take forever to check them out. I'm on Solaris, so there's no mtlib utility. I can run lbtest but that only lets me do queries. Server is TSM 3.7.4 on Solaris 7 Library is STK PowderHorn (9310) -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: ANR7804I mystery
Lesley - You don't specify, but I would assume that your system boot initiation files invoke the TSM server, Yes. The startup script claimed it didn't come up. which would account for the prior instance. Though that initial instance was either delayed or non-functional (you could not enter into an administrative session) the process itself would be there. In dropping the operating system to single user mode that should have disposed of the stifled initial process. Well, I thought I checked that. I created the startup script and don't entirely trust it yet so I'm pretty sure I did ps -ef|grep dsm to see if it was in fact already running. But it's not in my .bash_history so I guess I'll never know. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: AIX TSM scheduler problem
Richard Sims wrote: Yes, indeed. In Unix parlance, sh is the Default Shell, or Standard Shell - a rather non-specific reference to whatever actual shell the operating system developers deign to employ as their standard. In early AIX, it used to be Bourne shell. These days, in AIX and Solaris it is Korn shell. Not in Solaris. At least, not in 2.6 or 7. Instead, sh is linked to jsh (which I know nothing about). Solaris 7: tsmlibrary# cd /bin tsmlibrary# ls -l sh ksh jsh -r-xr-xr-x 3 bin root 91668 Oct 6 1998 jsh -r-xr-xr-x 2 bin bin 192764 Oct 6 1998 ksh -r-xr-xr-x 3 bin root 91668 Oct 6 1998 sh Solaris 2.6: nzwnsubo10001# cd /bin nzwnsubo10001# ls -l sh ksh jsh -r-xr-xr-x 3 bin root 88620 Jul 16 1997 jsh -r-xr-xr-x 2 bin bin 186356 Jul 16 1997 ksh -r-xr-xr-x 3 bin root 88620 Jul 16 1997 sh ASIDE A note for anyone wondering about Solaris version numbers, the more recent part of the sequence goes like this: 2.5 2.5.1 2.6 2.7 also known as 7 [2.]8 There is more to the story but this will do, I think. /ASIDE -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: Summary table
Maybe because copypool tapes are reclaimed under a single process. Where as primary pool volumes are reclaimed 1 volume at a time. Each its own process. No, that's not it - my copypool reclaim processes just don't show up at all. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: Antwort: Re: total # of bytes by each client on a daily basis
Gerhard Wolkerstorfer wrote: Do you know, what the value in the Field BYTES means ? When the activity is BACKUP, is the value in the field BYTES the amount of BYTES TRANSFERRED or BYTES BACKUPED ? William Boyer wrote: I'm assuming it's the amount of data transferred. And I say: I don't think it is. I have just made a comparison for several clients, and although for most of them the figure is pretty similar, there is one that shows a significant difference - the one that has large files to send, which won't fit in the diskpool. The summary table gives 323,035,314 and the actlog gives 355.94 MB which works out at 373,230,141. The difference is around 50MB or 15%. So, I think it's the number of bytes backed up. (And I think I'm going to change one of my scripts... in fact I think I can feel a diagnostic tool coming on...) -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: Starting Script from server console
Wolfgang Herkenrath wrote: My TSM-Server is Version 3.7.3.0 on OS390 R10. Here a extract from Administrators Guide: The scripts can be processed directly on the server console, the web interface, or included in an administrative command schedule. I tried the following command from MVS: /f tsm,run redologstandard I get the following message: ANR1491E Server command scripts cannot be started from the server console. So, is there a contradictory or am I stupid??? You are not stupid (as far as I can tell, anyway!) I can't find any similar statement in my admin guide, and I cannot run scripts from the server console either (3.7 on Solaris). I think it's a mistake in your admin guide, it should say that they can be processed from the command line admin client (or whatever you have in OS/390), not the server console. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: Change IP Address
Geoff Gill wrote: I need to change the IP address on my ADSM server this weekend. I want to know if there are any pitfalls I should look out for. It's not just an IP change on the same network, I'll be changing the IP, NETMASK and Default Gateway. Your clients need to know how to contact the server at the new address. You may already have that under control, but since you didn't mention it, I'll assume it slipped your mind. Do your clients' option files contain A. TCPSERVERADDRESS xxx.xxx.xxx.xxx, OR B. TCPSERVERADDRESS something.or.other.com ? - If A, then you need to update every client's option file and restart their schedulers - If B, then you need to make sure the DNS is changed, and that the changes will propagate to the clients within a reasonable time Contact your local DNS guru for more info. - If a mixture of A and B, then both apply, and also you may want to tidy up a little All of the above applies regardless of whether you're running in prompted mode or polling mode. (In prompted mode the A clients may or may not keep working after the chance, but they would fail the next time they were restarted.) I'll leave the other questions to those who actually know AIX. Questions: Since I'm not totally AIX literate yet is there a command line command I can use that will do all 3 at once? Should I plan on rebooting the box after this change?
Re: Change IP Address
Then again I could be forced to change the name in the future so what the hell.. If you do have to change the server name, perhaps you could use a CNAME, also known as a DNS alias. For instance my test server is called nzwnstsm10002.ems.itf.nz.eds.com (how's that for a mouthful!) there is an alias called tsm set up in the zone where the clients live so that the client option file has tcpserveraddress tsm. And in a production zone, the tsm alias points to the appropriate production server. This is Good Stuff because it means I can have the same option file everywhere, which makes it really easy to roll out more clients. It's also in keeping with traditional practice of having www pointing at web servers, ftp pointing at ftp servers and so forth. Anyway, good luck for the change. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Re: Reclaims Ques
Lindsay Morris wrote: How long does it usually take for the reclamation process to actually stop, and release the tape drive(s) so the restore can use them? I've typically seen this be 40 minutes or more, but I'm not sure why, or if something could be done to speed it up. I think it might depend on the size of the file the reclamation is processing currently and how much of it is left to go. It just so happens we've been doing some restore testing and reclamation at the same time on a 2-drive library, so I've looked back and the logs to see how long it has taken to finish a preempted process. So here are some actual real-life examples from here: ProcessVol dismount Process ended --- - Reclamation 3 min 5 min Reclamation n/a9 min Reclamation1 min 2 min Reclamation n/a 1 min Migrationn/a2 min Move datan/a3 min The reason for the missing volume dismount times would be because the restore wanted a volume that was already mounted. NB: this is on nasty 8mm tapes. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] I feel that there is a world market for as many as five computers Thomas Watson, IBM corp. - 1943
Deleting backupsets out of volhist (old chestnut, sorry)
Running 3.7.4 on solaris. I had three backupsets. One was deleted before it expired, the other two were allowed to expire. Now I have three tapes that I can't use. This was posted back in Feb... -Original Message- From: France, Don G (Pace) Subject: Re: I cant delete a backupset . . . There is a known APAR for this; it's fixed in 3.7.4 (for 3.7),,, don't know when it's fixed for 4.1. ...but in fact it's only half-fixed. There is no way to delete backupsets from the volume history. Is there any further progress with this? I can think of two workarounds: 1. Delete all volhist up to the date they were made. 2. Relabel the tapes with a new name. ...but I don't really want to use either of them because I'm using volhist to track how many times tapes have been re-used.
Re: MSCS and firewalls
Joel Cooper wrote: [much snippage] Does anyone have any ideas? Is this because my server has the option to be contacted on TCPPort 1500? I don't want to use the Windows scheduler to handle this, but I have to cover these servers. The server can only listen on one port - if you define it more than once in dsmserv.opt, it will use the last one. Are you running the scheduler in polling mode or prompted? The limited amount of testing I have done shows that when you use prompted mode, the server contacts the client on the assigned port, and all the traffic goes through that one server-initiated TCP session. This good because it means one less inbound port that you have to open up. I haven't done any tests with more than one client though, so I'd be interested to hear how you get on. I'm not currently running any clients through a firewall, but I'm a place where it could be required sometime. So I'm interested in this topic, but I just haven't done a lot of testing.
Behaviour during reclaim
I'm scratching my head trying to figure out why this is happening: I have a reclaim running, no other processes, no client sessions. Collocation is off. The reclaim is moving data from volume A to volume B, there is another volume C in the same stgpool with FILLING status, and volume D is scratch, unassigned. When B reaches end-of-volume, I would expect it continue by mounting C - but no, it takes the scratch tape, defines it in that stgpool and uses it - even though there would have been more than enough space on C to have taken all the data. Is there some reason why it does this? I would prefer it to leave the scratch volume alone so that it would be available to whatever stgpool might need it in the next 24 hours. Server is 3.7.4 on Solaris Library is an Exabyte 8mm thing defined as SCSI (2 drives, 10 slots) -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] "I feel that there is a world market for as many as five computers" Thomas Watson, IBM corp. - 1943
Re: WinNT Client Upgrade - Overwrite or Uninstall/Reinstall
For Solaris clients the upgrade method is to uninstall the current version (which leaves the relevant config files intact) then install the new version. I don't know for sure if the NT client is the same, but it seems likely. The Quick Start manual or the README for the new version should describe how to do the upgrade, but failing that, try out the uninstall-reinstall path on a non-critical client, and see how it goes. -Original Message- From: Scott Behrens [mailto:[EMAIL PROTECTED]] Sent: Friday, March 23, 2001 1:32 AM To: [EMAIL PROTECTED] Subject: WinNT Client Upgrade - Overwrite or Uninstall/Reinstall Please excuse if this is obvious - several archive searches turned up nothing, and I didn't see a reference in Richard Sims' Quick Facts. Relevant Installation Info: TSM Server 4.1.1.0 on AIX 4.3.3 WinNT Clients: TSM Client 4.1.1.0 on Intel NT 4.0 SP5 or SP6 Goal: Upgrade of WinNT TSM Clients from 4.1.1.0 to 4.1.2.12 to handle DST problem before April 1 (avoiding 4.1.1.16 because of potential memory-leak problem (IC28969)) If I install the 4.1.2.12 client over the top of the existing 4.1.1.0 client, then I end up with two 'Tivoli Storage Manager Client' entries in the 'Control Panel--Add/Remove Programs' (and duplicate entries for the two installed products in the Registry as might be expected). The 'Installing the Clients' manual, Version 4.1, Chapter 4. has a section, 'Reinstalling the Client' , which implies that if you are installing to the same location, you can just install over the top of the previous version with no uninstall required. I want to provide our NT sysadmins a straight-forward procedure for doing this upgrade, and don't know whether to suggest an uninstall-reinstall (ensuring that they have a backup of the dsm.opt and dsmsched.log), or just have them do an install over the top of the existing client (and put the footnote about the duplicate registry entries in the 'Book of Who Cares'.) Any thoughts on best practices? Thanks, Scott Behrens
Re: TSM 3.7 to 4.1 Upgrade
Shekhar Dhotre wrote: I have experienced this before , upgrade deletes all drive info , and if you try to reconfigure it , using smit tivoli ..you can`t you have to 1) delete drives from ODM use odmdelete .. (#rmdev -dl rmtx --does not work) and Geoff Gill wrote: I upgraded from 3.1 to 4.1 a month ago and did not have to do anything with drives. My system upgrade went very smooth, all devices that were there before the upgrade were there after it. I think there are two different situations here. Mark and Geoff (and I think others) wrote that they did not have this problem going from one major version to another (eg 3.7.x to 4.1.x) - whereas Wanda said "We just did a 3.7.2 to 3.7.4 upgrade..." and had the problem. Looks to me like the problem could be with 3.7.4 specifically. Shekhar, what version did you upgrade to that caused the problem? (not that it matters to me, I'm on Solaris - I just hate to see people arguing)
Re: SAN - FC Adapters doubt ...
Lesley Walker wrote: I have two FC64-1063 cards in an E4500, connecting to an EMC array. I installed the driver supplied by EMC (fcaw259). Sorry, forgot to mention, I have also installed patches as specified by EMC (back in November): 106541-12, 107458-10, 106924-06
Re: Attention Please [off-topic]
Suad wrote: loosely translated: We, the undersigned abhor the use of Afghani coffee beans in our expresso machines... Half a point! In case anyone REALLY wants to know, it's about the Taliban's mistreatment of women, and somebody wants to petition the United Nations to take action. Fabrizio: As much as I disapprove of the Taliban's abhorrent treatment of women, it is not an appropriate subject for this mailing list. If everyone posted messages here about every worthy cause, we would never get around to discussing TSM/ADSM.
Re: TSM Trivia Question
From: James Thompson [mailto:[EMAIL PROTECTED]] I thought it meant you could send an electric shock to the user at that machine Maybe I'll open a design change request to add that functionality. ;) Are you by any chance related to the BOFH?
Re: How to find out what tape's a client data are on ?
Andy Raibeck: Either way, I'm pretty sure that DISTINCT is supposed to account for distinct rows in their entirety, and not just certain columns. I'll have to look into this. That's what I thought too. My guess would be something to do with the SQL interface not being direct to the database - something going missing in the translation. Anyway, it's good to know I'm not going mad. :-) In case you need to know, the system that produces these results is 3.7.4 on Solaris, recently upgraded from an earlier 3.7 release.
Re: TSM, SAN, ACSLS and Solaris
I don't think the IFP driver is installed: tsm22001# modinfo | grep -i ifp tsm22001# But thanks for the answer. I've got the st.conf sorted - I had done that before on a different box but it was directly-attached FC-AL rather than fabric. It's more the question of what to put in mt.conf and when to do "add_drv mt". And thanks to Suad and Richard for your answers too. I now feel a lot better equipped to work with the storage guys. I always get neverous when someone says "just give me the root password, I'll configure it" - especially when I'm the one who has to produce the documentation. -Original Message- From: George Lesho [mailto:[EMAIL PROTECTED]] Sent: Friday, March 02, 2001 5:19 AM To: [EMAIL PROTECTED] Subject: Re: TSM, SAN, ACSLS and Solaris To configure the qlogic hba, did you first remove the Sun IFP driver? You probably need to do this and reinstall the qlogic driver with the vendors id#'s * rem_drv ifp * rem_drv qla2200 * add_drv -c scsi ' "vendor's devide id #" ' qla2200 * Then you will edit your st.conf file to make sure support for the tape drives you are using is selected by uncommenting references to them This will look like: tape-config-list= "QUANTUM DLT7000","Quantum DLT7000", "DLT7K-data"; DLT7k-data =1,0x38,0,0x1D639,4,0x82,0x83,0x84,0x85,2; there is some sort of readme file to explain how to link the wwn with the devices as I recall.. I haven't seen a swith or FC for a couple years... If I recall correctly this will be done in the qla2200.conf file. * touch /reconfigure * reboot - - -r The system reads the driver config file as well as the st.conf and builds NEW device files in /dev/rmt... in fact, you can remove the old ones first if you like, they won't work if the stuff isn't configured right anyway. You will see references to the new HBA your banner and /var/adm/messages once it is installed right. If you have syntax errors in the st.conf, these will show up in the banner and /var/adm/messages. Does qlogic or the O/S (sorry I am in an all AIX env now) give you a utility called drvconfig? If so, I think you can also run the thing to build your special dev files... drvconfig -i qla2200 devlinks tapes ---then look into /dev/rmt to see if the special character files necessary for dev support were created George Lesho Storage/System Admin AFC Enterprises drvconfig? "Walker, Lesley R" [EMAIL PROTECTED] on 02/28/2001 09:24:50 PM Please respond to "ADSM: Dist Stor Manager" [EMAIL PROTECTED] To: [EMAIL PROTECTED] cc:(bcc: George Lesho/Partners/AFC) Fax to: Subject: Re: TSM, SAN, ACSLS and Solaris Hi Ron, thanks again for the answer. Sorry to be asking so many questions, I just don't know much about fibre-channel and even less about switched fibre-channel. So, is a world wide name a sort of alias that creates a tunnel through the switch in this case? Could I just pick, say SCSI-ID 2 and put in a line that says hba1-SCSI-target-id-2-fibre-channel-name="something" and then as long as "something" points to the right thing it will work? Or does it have to be the right SCSI-ID? One thing that's holding me back is not having any doco for the Qlogic card (it was "borrowed" from another project) so thanks for pointing me at the qla2200.conf file. Another general Solaris question, is it safe to do "boot -r" before running the add_drv commands? I was told very early on to never do boot -r at all, but I suspect this advice may have been a bit bogus. We have JNI cards in the box as well (for connecting to EMC) so it gets rather confusing! (The box is an E4500 if that matters) -Original Message- From: Ron Pavan [mailto:[EMAIL PROTECTED]] Sent: Thursday, March 01, 2001 12:36 PM To: [EMAIL PROTECTED] Subject: Re: TSM, SAN, ACSLS and Solaris I believe what you are looking for is placed in a conf file. example: We have a sun e3500 running with QLogic HBA's. We define a target-id to a world wide name on a fiber switch (can can also define a target to a port number on the switch but then you are limited to that port and if the port goes bad you cannot switch ports without a reboot). This config files is \kernel\drv\qla2200.conf. When you reboot/reconfigure the st.conf should scan for targets that can now be found because the HBA is defining them. -Original Message- From: Walker, Lesley R [mailto:[EMAIL PROTECTED]] Sent: Wednesday, February 28, 2001 2:34 PM To: [EMAIL PROTECTED] Subject: Re: TSM, SAN, ACSLS and Solaris Thanks for the answer, and thanks to the others who answered too. The part that I'm trying to figure out now is how to install the drivers. With Solaris, you have to edit a file that contains definitions of the SCSI-IDs of the tape drive. How can that work if the tape drives are assigned via a switch? Do you set up some kind of permanent virtu
Re: How to find out what tape's a client data are on ?
Just out of curiosity, since I've only casually been following this thread, was there a specific reason for the "group by" clause? I think what is desired is simplly an "order by" clause, e.g. Well, I'm really confused now. I came up with my own version of the query: select distinct volume_name,stgpool_name from volumeusage where node_name='FOO' and that produces an answer: VOLUME_NAMESTGPOOL_NAME -- -- ITF010 CPITFPOOL ITF013 CPITFPOOL ITF018 CPITFPOOL ITF026 CPITFPOOL ITF001 ITFPOOL ITF003 ITFPOOL ITF004 ITFPOOL ITF014 ITFPOOL ITF025 ITFPOOL ITF027 ITFPOOL ITF028 CPITFPOOL ITF006 ITFPOOL ITF002 CPITFPOOL BUT if I add "order by stgpool_name" it doesn't show all the tapes - eg: VOLUME_NAMESTGPOOL_NAME -- -- ITF010 CPITFPOOL ITF001 ITFPOOL So what's going on here?
Re: TSM, SAN, ACSLS and Solaris
Thanks for the answer, and thanks to the others who answered too. The part that I'm trying to figure out now is how to install the drivers. With Solaris, you have to edit a file that contains definitions of the SCSI-IDs of the tape drive. How can that work if the tape drives are assigned via a switch? Do you set up some kind of permanent virtual circuit so that you always get the same drives? Please excuse my ignorance - I have no access to the SAN switch doco, as it is being set up by people in another team in a different city. The latest information I have on the switch is that it is either Inrange FC9000-16 or Brocade Silkwork 2800. -Original Message- From: Collins, Brenda [mailto:[EMAIL PROTECTED]] Sent: Wednesday, February 28, 2001 1:11 AM To: [EMAIL PROTECTED] Subject: Re: TSM, SAN and ACSLS We are using the STK Powderhorn with 9840 drives in a TSM 3.7 (AIX) environment and TSM4.1 (OS/390) environment. We will soon be implementing some fiber drives also to a new TSM (AIX) server. Whenever you use the STK libraries, you are forced to use the ACSLS software also. We do have a tape SAN implemented. Brenda -Original Message- From: Walker, Lesley R [mailto:[EMAIL PROTECTED]] Sent: Monday, February 26, 2001 9:52 PM To: [EMAIL PROTECTED] Subject: TSM, SAN and ACSLS Sorry if this is an FAQ, I've spent a bit of time looking but haven't found a definitive answer. Is this possible: TSM 3.7.x on Solaris connected to SAN switch (don't know what model, but assume it's a supported one) connected to StorageTek PowderHorn (9840 drives, ACSLS controlled) From a quick look at the Redbook (Using Tivoli Storage Manager in a SAN environment), SCSI libraries are supported from TSM 3.7 onwards, but I can't find any mention of ACSLS. Has anyone done it? Tried it and failed? This is the configuration I am expected to implement, and I don't have much experience of SANs, so any hand-holding would be appreciated. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] "I feel that there is a world market for as many as five computers" Thomas Watson, IBM corp. - 1943
Re: TSM, SAN, ACSLS and Solaris
Hi Ron, thanks again for the answer. Sorry to be asking so many questions, I just don't know much about fibre-channel and even less about switched fibre-channel. So, is a world wide name a sort of alias that creates a tunnel through the switch in this case? Could I just pick, say SCSI-ID 2 and put in a line that says hba1-SCSI-target-id-2-fibre-channel-name="something" and then as long as "something" points to the right thing it will work? Or does it have to be the right SCSI-ID? One thing that's holding me back is not having any doco for the Qlogic card (it was "borrowed" from another project) so thanks for pointing me at the qla2200.conf file. Another general Solaris question, is it safe to do "boot -r" before running the add_drv commands? I was told very early on to never do boot -r at all, but I suspect this advice may have been a bit bogus. We have JNI cards in the box as well (for connecting to EMC) so it gets rather confusing! (The box is an E4500 if that matters) -Original Message- From: Ron Pavan [mailto:[EMAIL PROTECTED]] Sent: Thursday, March 01, 2001 12:36 PM To: [EMAIL PROTECTED] Subject: Re: TSM, SAN, ACSLS and Solaris I believe what you are looking for is placed in a conf file. example: We have a sun e3500 running with QLogic HBA's. We define a target-id to a world wide name on a fiber switch (can can also define a target to a port number on the switch but then you are limited to that port and if the port goes bad you cannot switch ports without a reboot). This config files is \kernel\drv\qla2200.conf. When you reboot/reconfigure the st.conf should scan for targets that can now be found because the HBA is defining them. -Original Message- From: Walker, Lesley R [mailto:[EMAIL PROTECTED]] Sent: Wednesday, February 28, 2001 2:34 PM To: [EMAIL PROTECTED] Subject: Re: TSM, SAN, ACSLS and Solaris Thanks for the answer, and thanks to the others who answered too. The part that I'm trying to figure out now is how to install the drivers. With Solaris, you have to edit a file that contains definitions of the SCSI-IDs of the tape drive. How can that work if the tape drives are assigned via a switch? Do you set up some kind of permanent virtual circuit so that you always get the same drives? Please excuse my ignorance - I have no access to the SAN switch doco, as it is being set up by people in another team in a different city. The latest information I have on the switch is that it is either Inrange FC9000-16 or Brocade Silkwork 2800.
TSM, SAN and ACSLS
Sorry if this is an FAQ, I've spent a bit of time looking but haven't found a definitive answer. Is this possible: TSM 3.7.x on Solaris connected to SAN switch (don't know what model, but assume it's a supported one) connected to StorageTek PowderHorn (9840 drives, ACSLS controlled) From a quick look at the Redbook (Using Tivoli Storage Manager in a SAN environment), SCSI libraries are supported from TSM 3.7 onwards, but I can't find any mention of ACSLS. Has anyone done it? Tried it and failed? This is the configuration I am expected to implement, and I don't have much experience of SANs, so any hand-holding would be appreciated. -- Lesley Walker Unix Engineering, EDS New Zealand [EMAIL PROTECTED] "I feel that there is a world market for as many as five computers" Thomas Watson, IBM corp. - 1943
Re: TSM and Restore....
Etienne Brachel wrote: Thanks for your suggestion but for 40 clients using collocation on filespaces will (as we use a diskpool) give us a lot of migration problems having only 4 drives available. You would need to create a separate storage pool hierarchy for that node. Or do you have similar amounts of storage for your other clients? If so I have now run out of suggestions.
Re: TSM and Restore....
For your current restore, I think you are pretty much stuck with the 266 mounts. To make your life easier in the future, you could update the stgpool to collocate=filespace, then it will try to keep separate tapes for each filesystem. The change would only apply to new volumes in that storage pool, so you might want to do a "move data" on all of the existing volumes - it would depend on how quickly it expires. -Original Message- From: Etienne Brachel [SMTP:[EMAIL PROTECTED]] Sent: Thursday, January 18, 2001 9:06 AM To: [EMAIL PROTECTED] Subject: TSM and Restore Hi TSM'rs, I had a strange experience today. I worked on a Sun Solaris system and it has the client 3.7.X installed. The system is backed up for approx. 5 years now A library is used containing 4 drives. This system has about 130 GB of data and collocation is used to keep data as much as possible on same volumes. In total there are 19 Volumes !! We needed to make a script to restore all filesystems to the Solaris machine. because "dsmc res -subdir=yes -preservepath=yes "/ /var /opt" "/restore/"" is not possible. Only one source, for example "/var", can be specified. --ISSUE-- Why isn't it possible to start one command-line session to restore all filesystems ? All restores are seperate sessions (filesystem by filesystem) . Every Session needs data of all 19 tapes (Incremental Forever). Why does TSM mount all 19 tapes for each session ? We have about 14 filesystems that need to be restored. RESULT : 266 Tape Mounts..!! How does the GUI handle this ? does it also start a seperate session for each filesystem or directory that was selected for restore ? OR does it collect info for restore to see that only a one time mount is necessary to read the data of a volume for all filesystems that need to be restored ? Am I doin anything wrong ? or is this how restores are treated with command-line sessions... or maybe an overall treatment.. ? Is there another way of doin this ? because a 14 hours restore is necessary to restore all filesystems with collocation enabled... I think that is very poor.. Hope someone can help me... Thanks and regards, Etienne Brachel Touch The Progress Services b.v. Certified Tivoli Storage Management Consultant [EMAIL PROTECTED] _ Get Your Private, Free E-mail from MSN Hotmail at http://www.hotmail.com.
Re: TSM 4.1.2, SUN Solaris, StorEdge L20
Viliam Varga wrote: TSM device driver (lb) cannot identify library contrloler on tape library SUN StorEdge L20 (2 drives, 20 tapes). TSM tape device driver (mt) works fine on both DLT 7000 drives. Veritas NetBackup BusinesServer 3.3 software installed on the same machine uses library robot and drives without failures. This sounds remarkably similar to the problem I had with my StorageTek L20 - I assume it's the same thing, since the description matches. I was trying to use it with TSM 3.7.3, and it didn't work until I upgraded the code to 3.7.4. (I posted to the list about it at the time, so you should be able to find details in the list archive if you want them.) But the other thing I did was re-install the OS in 32-bit mode. It could be worth a try - and if you contact Tivoli about it, they will very likely suggest doing it. If that doesn't work, my suggestion is to log a call with Tivoli. They made some kind of change between 3.7.3 and 3.7.4 that made it work for me - so it's possible you might need the same change at the 4.1 level. CONFIGURATION: Hardware: SUN Ultra5, 256MB RAM, 16GB HDD SUN StorEdge L20, 2x DLT7000 Software: Solaris 2.7, 64bit TSM server 4.1.2 Thank you. Viliam Varga
Re: backup of /tmp on AIX 4.3.2, TSM client 4.1.1
Richard Sims wrote: Try: dsmc incremental -subdir=yes /tmp/sumpin/ That works w/ AIX 4.3.2 ADSM client v3.1.0.7! Is the TSM client v4.1.x "Broken, As Designed"? It doesn't work on the AIX TSM 3.7.2.0 client. Doing 'dsmc show inclexcl' yields: Excl All /tmp/.../* despite there being no /tmp Exclude specified in customer files. Adding /tmp to the DOMain option makes no difference. A look in the APARs database shows confusion and disagreement in Tivoli as to what should occur. The Unix Clients manual doesn't say anything about /tmp outside the context of ALL-LOCAL, and is vague there. Sigh. It works for me with the 3.7.2.0 client on Solaris (see following). I tried it with two different server versions too (3.7.0 and 3.7.4), and I haven't even thought about using a domain statement. Maybe you have /tmp excluded in a client option set on the server? If not, then there is a difference in behaviour between Solaris and AIX. # dsmc Tivoli Storage Manager Command Line Backup Client Interface - Version 3, Release 7, Level 2.0 (C) Copyright IBM Corporation, 1990, 2000, All Rights Reserved. tsm show inclexcl Node Name: NZWNSTSM10002 Session established with server TSMLIBRARY: Solaris 2.6/7 Server Version 3, Release 7, Level 0.0 Server date/time: 01/15/2001 11:57:17 Last access: 01/15/2001 11:55:13 Mode Function Pattern (match from top down) - - Excl Filespace /tivoli No exclude directory statements defined. No include/exclude statements defined. tsm i -subdir=y /tmp/foo/* Incremental backup of volume '/tmp/foo/*' Directory-- 113 /tmp/foo/bar [Sent] Normal File-- 485 /tmp/foo/foo-file [Sent] Normal File-- 712 /tmp/foo/bar/foobar-file [Sent] Successful incremental backup of '/tmp/foo/*'
Re: TSM security - sharing STK library with other apps
Thanks for the replies everyone. Oddly enough I hadn't thought of the possibility of the other app stomping on TSM, that's what comes of being too busy I guess. I think the other app is Netbackup, does anyone know how it might behave in this context? -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949
Re: node is missing its scheduled event but daemon is running
There are two possible client scheduling modes, polling and prompted. Your client is in "prompted" mode. The server can be set to allow "polling", "prompted", or "any". Do a "q stat and look for this line: Scheduling Modes: Any If it is set to polling, that will be why your client has missed its schedule. To put the client back to polling mode, go to its options file and change it from "schedmode prompted" to "schedmode polling", and restart the daemon. If you want to change it on the server, the command is "set schedmode any". -Original Message- From: Reinhold Wagner [SMTP:[EMAIL PROTECTED]] Sent: Friday, December 22, 2000 6:21 AM To: [EMAIL PROTECTED] Subject: node is missing its scheduled event but daemon is running folks, one of our *SM client nodes (TSM v4) is missing its schedule since monday. In dsmsched.log we find after _next operation scheduled:_ the following message: Waiting to be contacted by the server. while our v3 clients show this message: 12/21/00 14:49:26 Server Window Start: 22:20:00 on 12/21/00 12/21/00 14:49:26 12/21/00 14:49:26 Schedule will be refreshed in 1 hour. Server is 3.1.2.40. Seems that the missing node is waiting to be triggered by the server instead of querying the schedule and start the backup by itself. Any ideas? TIA Reinhold Wagner, Zeuna Staerker GmbH Co. KG
TSM security - sharing STK library with other apps
Does anyone have experience of using an ACSLS-controlled STK library and sharing it with other applications? Our customer is very concerned about the possibility of Application A being able to read/overwrite tapes belonging to Application B, and the question has been asked: Most (all?) tape management systems have the ability to automatically load the entire ACSLS database into their own database and this is where the main risk arises. Does TSM have this ability? They will be implementing access control, but it's not in place yet. Can I assure them that TSM will not be a security risk? (Version 3.7.3 on Solaris) -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949
Re: StorageTek L20 (outcome)
Yes, right from the beginning when I first plugged it in and turned it on, it showed up in the results of the probe-scsi-all. -Original Message- From: Shekhar Dhotre [SMTP:[EMAIL PROTECTED]] Sent: Wednesday, November 29, 2000 3:15 AM To: [EMAIL PROTECTED] Subject: Re: StorageTek L20 (outcome) Does probe-scsi-all shows STk library ? I am using STK9710 , with SUN no problem .. "[EMAIL PROTECTED]/P=Internet/A= /C=us" on 11/27/2000 10:50:16 PM Please respond to "[EMAIL PROTECTED]/P=Internet/A= /C=us" @ X400 To: "[EMAIL PROTECTED]/P=Internet/A= /C=us"@X400 cc: Subject: Re: StorageTek L20 (outcome) Just in case anyone's interested in the outcome, I've finally got this library working. The suggestion from Tivoli was to have the operating system in 32-bit mode, since other people with the same problem also were running in 64-bit mode, and also to upgrade to the 3.7.4 server code. First I tried booting in 32-bit mode (by doing "boot kernel/unix" at the ok prompt) but it didn't help. So then I rebuilt the server from scratch in 32-bit mode, and it still didn't help. Finally I upgraded to the 3.7.4 server code, and that did the trick. Moral of the story: if you want to use an L20 library with Solaris, you need to be at 3.7.4 or better.
StorageTek L20
Hi all, I need help to get my tape library going. It's a StorageTek L20, connected to a server which is TSM 3.7.2 on Solaris 7. I've connected up the cables and done a boot -r (to get the OS to recognise the new devices, which it appears to). But when I try to define the library, the admin client just hangs. Are there any OS tools I can use to test the library? More details: Robotics are accessed via SCSI - it should be able to work the same as any other SCSI library. Robotics and drive are on the same SCSI chain. Library is target 0, drive is target 1 Cabling goes server-library-drive-terminator -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949
Re: StorageTek L20
It's a StorageTek L20, connected to a server which is TSM 3.7.2 on Solaris 7. Sorry, that's TSM 3.7.3 (was getting confused with clients) I've connected up the cables and done a boot -r (to get the OS to recognise the new devices, which it appears to). But when I try to define the library, the admin client just hangs. Are there any OS tools I can use to test the library? More details: Robotics are accessed via SCSI - it should be able to work the same as any other SCSI library. Robotics and drive are on the same SCSI chain. Library is target 0, drive is target 1 Cabling goes server-library-drive-terminator -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949
Re: Client ver 3.7.2 - no SA software
Can't speak for the admin gui because I've never seen it (starting out on 3.7, never seen any earlier versions), but web admin isn't totally useless. There's one thing in particular that's vastly easier with web admin than with the command line: writing and updating scripts. But so far, I prefer the command line for most things. -Original Message- From: Fred Johanson [mailto:[EMAIL PROTECTED]] Sent: Friday, November 10, 2000 6:21 AM To: [EMAIL PROTECTED] Subject: Re: Client ver 3.7.2 - no SA software I've also kept my gui alive for these reasons, plus the ability to filter requests and backtrack. Changing copygroup settings in a domain is a few clicks on a limited list in the gui with a much easier return to your starting point. The only fans of the WebAdmin around here are those who have never used the gui.
Re: How to move data from error tape
Hi friends, Has anyone experience a write tape error after a number of percent doing a backup particularly schedule backup, then it will continue on another scratch tape until completes. How can i move the data on the write error tape to a new scratch tape (to avoid any problems with the rite error tape when restoring) with only 1 drive available? I haven't actually had to do this (yet) but I think you would have to move it to a disk storage pool and back again - which means you need an empty disk pool to work with. Also with regard to your other question about how to do reclamation with only one drive: There is a parameter called Reclaim Storage Pool which is used for this purpose - again, the data is moved to a diskpool and then back to a new tape. So what you need to do is set up a diskpool, preferably one that's big enough to hold the data from a full tape, and then update your tape pool with reclaimstg=diskpoolname. Again, this is just theory from me, I have not done it in real life (but I will have to do these things soon).
Re: Restore Sun Solaris from scratch
I am also interested in answers to this, as it is something I will eventually have to address too. Obviously you cannot restore without installing the OS to run the client, and having the right filesystem layout to restore into. This means that there are two basic options: 1. install the OS to a different disk, set up the filesystems, and then you have to know how to restore to a different location (because the filesystems will not be mounted with their normal names) 2. identify which OS files you cannot safely restore while the OS is running, and exclude them from backup. There could be variations on #1, such as creating a boot CD with the OS and client on it. With regard to #2, does anyone have a list of such files? -Original Message- From: Arturo Lopez [SMTP:[EMAIL PROTECTED]] Sent: Thursday, November 02, 2000 10:13 AM To: [EMAIL PROTECTED] Subject: Restore Sun Solaris from scratch Hello All Is anyone familiar with restoring Sun Solaris from scratch. I have a Sun Sys Admin that does not believe TSM can restore the entire OS from scratch. With AIX we use a SYSBACK to take an image of the OS and restore the image and then apply the incremental to bring the OS back to pristine level. Does anyone have experience restoring a Sun Solaris box from scratch... On the Intel world running WinNT 4.0. We load a second instance of NT and then boot into the second instance and restore the original OS and Registry. Then reboot server into original instance.Can you do this with Sun.. Thx Arturo Lopez IT Systems Programmer 210.913.1845 210.753-1845
Running with only one drive
Due to circumstances that I won't bore you all with, I am going to have to implement a small TSM server with a DLT auto-changer that has only one tape drive. This is to a be temporary situation but I want it to be sustainable "just in case". The total amount of data to be backed up is easily able to be fitted into the disk space that I have available, so my plan is to have all primary storage in one diskpool, and a copypool on tape. Do I need to worry about the diskpool space becoming fragmented as old data expires? What's the best way to set up the diskpool - should I create multiple filesystem volumes in the space, or use raw disk volumes, or does it make no real difference? Would it be better to use a standard random-access diskpool, or should I create a sequential-on-disk pool? Performance is not much of a concern due to the small amount of data, I just want to make sure it can be done. The specifics are: 3.7.3 on Solaris 9GB (total) to be backed up from multiple clients (not expected to grow at all) 2 versions 4 x 9GB disks for diskpool -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949
Re: Platforms for TSM
Howard Heitner wrote: Currently, we are running TSM 3.7 on MVS. Since we are phasing out this platform, I am looking to move TSM to either SUN (Solaris) or IBM (AIX). We have good expertise on Solaris and very little on AIX. Also, we are using an IBM 3494 Tape Library with 3590-B1As. My question are: What are users experiences with running TSM under Solaris and AIX? Is support better under AIX then with Solaris? There are annoying things such as documentation that says "for NT do this... and for unix, do this..." and you find that the "unix" instructions are specific to AIX and need to be adjusted for Solaris. But so far I haven't found anything serious enough to make me want to switch to AIX. Given that you have Solaris expertise and not much AIX expertise, and all other things being equal, my personal opinion is go for Solaris. I should point out though, I'm biased because I have no AIX experience whatsoever. I'm also very new to TSM.
Re: Speaking of auto-replies
Hmm... I could have come up with a translation at least as good as that one, and I don't even know any German. :-) -Original Message- From: Alex Paschal [SMTP:[EMAIL PROTECTED]] Sent: Wednesday, October 11, 2000 1:49 PM To: [EMAIL PROTECTED] Subject: Re: Speaking of auto-replies Heh, heh. http://world.altavista.com has a translator that almost does the job. It helps me decide if I want a better translation, anyway. Here's what it has to say. Needless to say, I don't think it needed a better translation. "I have vacation and am on 13.10. again in the office. They can turn however gladly on mine colleague Heinz board bretthauer (mailto:bretthauerh@alte leipziger.de) or on mine boss refuge Harnischfeger (mailto:harnischfegerh@alte leipziger.de). Thank you for your understanding. Andreas rensch P.S. Their Mail was not passed on automatically. " Alex Paschal Storage Administrator Freightliner, LLC (503) 745-6850 phone/vmail (503) 745-5091 fax
Re: Option file and optionset
Michel Engels wrote: You have to combine both syntaxes. an example would look like: DEF CLIENTO FRED INCLEXCL EXCLUDE "/foo/*" Well I thought this looked hopeful, but no such luck. It still complains about EXCLUDE as an invalid keyword. The full output is as follows: tsm: TESTSERVERdef cliento inclexcl exclude /kernel/genunix ANR2056E DEFINE CLIENTOPT: Invalid option name - EXCLUDE. Retry using the full optionname. ANS8001I Return code 3. "Walker, Lesley R" [EMAIL PROTECTED] on 10/09/2000 06:16:26 AM While we're on this topic, can someone give an example of the syntax (3.7.3 on Solaris) to specify an include/exclude statement in the client option set? I've read the Admin Ref, and tried a couple of ways but it comes back with errors: DEF CLIENTO FRED EXCLUDE /foo/* results in: Invalid option name - EXCLUDE" DEF CLIENTO FRED INCLEXCL "/path/filename" DEF CLIENTO FRED INCLEXCL "ex /foo/*" both result in: DEFINE CLIENTOPT: Invalid option value (and yes, the file exists in the first one) What am I doing wrong? Why doesn't the manual have an example, dammit?
Re: Option file and optionset
Ron Pavan wrote: example in my env: all server side option sets for "excludes" are set with override = no all server side option sets for "includes" are set with override = yes [etc] While we're on this topic, can someone give an example of the syntax (3.7.3 on Solaris) to specify an include/exclude statement in the client option set? I've read the Admin Ref, and tried a couple of ways but it comes back with errors: DEF CLIENTO FRED EXCLUDE /foo/* results in: Invalid option name - EXCLUDE" DEF CLIENTO FRED INCLEXCL "/path/filename" DEF CLIENTO FRED INCLEXCL "ex /foo/*" both result in: DEFINE CLIENTOPT: Invalid option value (and yes, the file exists in the first one) What am I doing wrong? Why doesn't the manual have an example, dammit?
Re: Upgrading from ADSM 3.1.2.50 to TSM 3.7.x
Cathy Bowman wrote: Specifically, what are some of the problems with 3.7 besides those just mentioned? We've decided against going to 4.1 because its so new, but hearing this... I've found two things so far (Solaris platform) 1. Missing startup/shutdown scripts for dsmserv 2. Missing/incorrect startup/shutdown scripts for ACSLS But we're not fully operational yet, so there could be problems I haven't found.
How long does a HALT take?
I suppose it's like asking "how long is a piece of string", but, here's the question: I'm writing a Solaris startup/shutdown script that will do a halt and then wait in a loop for the dsmserv process to go away. This obviously needs a timeout in case the halt doesn't work. Currently, I only have one client registered and the database is tiny, and the halt takes about 5 seconds to complete. With a large database, will the halt take longer? If there are sessions active, will it take longer? How long should one reasonably wait for the server to finish halting? Would 60 seconds be enough? Any real-life examples would be greatly appreciated. -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949
Re: BACKUP DB to disk - questions
Richard Sims [SMTP:[EMAIL PROTECTED]] wrote: [various answers] Hi Richard, thanks for taking the time to answer, and thanks for the sample shutdown script, I'm sure I can do something similar for my server. It's a pity the startup/shutdown processes are so lacking in the product itself, hopefully Tivoli are working to improve on this. The best cause of db mangling is in your computer, opsys, or TSM suddenly going away, without graceful shutdown, with just a single image of your db. Use TSM db mirroring, with MIRRORWrite Sequential to stay out of trouble. Thanks for the hint, I see the defaults are sequential for the DB and parallel for the log. Would you recommend changing the log to sequential as well, or leave it as is? Normally I'm inclined to trust defaults unless I know better, but certain... experiences... have make me wary. Server programming errors also result in "interesting" situations, Hopefully this won't occur, as we have a policy of testing all changes in the lab before applying them to production.
Re: BACKUP DB to disk - questions
Prather, Wanda [SMTP:[EMAIL PROTECTED]] wrote: [Prather, Wanda] I agree with you that in an EMC Symmetrix, the liklihood of physical failure is very very small. I've personally never had to restore a TSM DB due to software corruption (and I'm not particularly gentle with mine), but other people on this list have. AND there is the possibility of human error to be paranoid about. Overall likelihood very small, but impact huge Thanks for taking the time to answer. It did occur to me, after I sent the message, that the most likely scenario for losing the EMC is if someone kicks the cable. :-) So a mirror on local disk seems like a good idea.
BACKUP DB to disk - questions
TSM 3.7.3 on Solaris 7, new implementation. command: backup db devclass=seqdisk Until my organisation gets its act together to give me access to the ACSLS server (it's a long story), I have no tape storage available. But I have a nice large chunk of EMC Symmetrix array to play in. So I have created a devclass with devtype=file, and I have used that as the destination for BACKUP DB, and it works just fine, and I'll go ahead and set up a schedule for it. But I do have some questions. Does TSM do anything about managing the database backup volumes? Or do I need to do that manually? (ie delete old versions before it runs out of space). As far as I can tell, the database doesn't see them as volumes, because they don't show up when I do Q VOL (although Q VOLHIST describes them as volumes. When I have to restore the database at a later date, will I actually need the volumehistory file? I figure I will be able to tell by the timestamps on the file, and the size, which one to use. And some more general questions: How likely is it that we'll need to restore the database anyhow - how often does a TSM database become corrupted, and under what circumstances? I'm assuming the media is 99.999% safe because it's in the Symmetrix array, in a mirrored configuration. The active database is in the same EMC, on a different LUN, also mirrored. So I'm betting that 2 x 2 mirrors are not going to fail all once. Barring the natural disaster scenario, is this adequate, or should I get more paranoid? I know these things are a trade-off between likelihood and impact - what I'm after is the size of the likelihood. Anyone have any experience they'd be willing to share? (Apologies for these newbie questions - I admit I'm a newbie at this, and there will be more. Feel free to point me at relevant docs - I have looked but there is SO much of it.) -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949
Re: stk 9840
Bill Colwell [SMTP:[EMAIL PROTECTED]] wrote: We got 9840's very early and 3590 emulation wasn't available. At the time we were running adsm 3.1 tsm 3.7 didn't exist, so there was no performance difference. You should choose 3590 emulation if there are no strong reasons for doing 3490 emulation. If your site is os/390 and has dfsms-hsm, you definitely want 3590 for block count reasons. Sorry, I guess I didn't frame my question clearly enough. I'm planning on using them in their standard STK9840 mode, and was wondering what was the point of emulating other tape types. My servers will all be running on Solaris and your mention of OS/390 is probably enough of a clue for me. That'll teach me for having OS-tunnel-vision.
stk 9840
Bill Colwell [SMTP:[EMAIL PROTECTED]] wrote: TAPEIOBUFS came in with 3.7. All my storage pools are on stk 9840s but they are emulating 3490's. I have tapeiobufs set to 9 anyway; after the storage pools were in use, we got some 9840's in 3590 mode. I hoped that I could do the dbb to the 3590s and it would be faster but it wasn't. Just out of curiosity, why do you have the 9840s emulating 3490s and 3590s? Is there an advantage to doing this, or is it just for compatibility with pre-existing tapes? The reason I ask is because I am in the throes of setting up a new implementation with STK 9840s (in a Powderhorn) and I don't know a lot about tapes yet.
How to add advanced device support?
I'm working on a new implementation, with no prior experience of TSM or ADSM (but I've done a training course!). Version: TSM 3.7.3 Platform: Sun E450 Library: StorageTek PowderHorn 3910 Deadline: yesterday (isn't it always?) To get the Powderhorn to do anything, I need to get the TSM server talking to the ACSLS server which provides the control path for the tape silo. To access the ACSLS from TSM, I understand I have to have "Extended device support", which seems to be also known as "advanced device support". We seem to have a licence for it. Can anyone help me with the process for getting and adding "advanced device support"? I assume it's something I haven't got or haven't done, because Q STAT tells me: TSM Is Advanced Device Support required ?: No TSM Is Advanced Device Support licensed ?: Yes The admin guide tells me "When you install TSM, you must choose whether to install the TSM device drivers for tape and tape autochanger devices." Someone else (now unavailable) did the installation, and I suspect he said no when he should have said yes. So what I'm looking for is a way to add advanced device support without having to trash the installation and restart from scratch. I've looked in a number of documents and not found the answer, maybe I'm not looking in the right place. Any assistance will be most welcome. -- Lesley Walker Distributed Systems Services, EDS New Zealand [EMAIL PROTECTED] "Where a calculator on the ENIAC is equipped with 18,000 vacuum tubes and weighs 30 tons, computers in the future by the year 2000, may have only 1,000 vacuum tubes and weigh only 1.5 tons" Popular Mechanics, March 1949