Re: Migration question
On 02/10/2007, at 2:35 AM, Dollens, Bruce wrote: I have a question that I feel a little stupid in asking. What is the difference between migration and backup primary (disk) to copy (tape)? Other people have already answered, so I won't bother. :) I am working on changing my scheduling up and the recommended order of steps I was given is: Backup clients Migration Backup primary to copy Backup db Expiration Reclamation Start all over again I'd change that order slightly to: Backup clients Backup disk pool(s) Migration Backup tape pool(s) Backup db Expiration Reclamation Repeat ad nauseum. Better to backup the storage pool data while it's still on disk, rather than to flush it out to tape and then back it up - saves on the amount of data read from tape. Some data may well still flush down to tape before the first backup stgpool runs, which is why you want to backup the tape pool as well after migration is complete.
Re: Possible to delete folders within a filespace?
On 04/09/2007, at 10:34 PM, Angus Macdonald wrote: Many thanks Richard. That looks like it'll do what I need. Can anyone point me to a redbook or other document that covers operation of the client command-line? I can't seem to find one anywhere. I'm looking right now at the IBM Tivoli Storage Manager for Unix: Backup-Archive Clients Installation and User's Guide. Possibly the wrong platform, but you should be able to extrapolate from there. (Available in both HTML and PDF form from IBM's website ... somewhere.) Cheers, Stuart.
Re: prescheduled postcheduled commands help
On 23/08/2007, at 12:29 AM, Głębicki Jakub wrote: Hint :) use short version of path (Progra~1 instead of Program Files) in P*SCHEDULECMD options, otherwise schedule will fail. A small trap for young players: the short name is defined by the order of directory creation. So if the server dies and you recover from backup, anything that relies on the short name is likely to fail if the recovery doesn't create the directories in exactly the right order. Isn't Microsoft *wonderful*? I don't know this from personal experience, but I have heard it from somebody whose experience and knowledge in these matters I trust. Verify if it's a concern for you.
Re: Why virtual volumes?
On 23/08/2007, at 7:29 AM, Nicholas Cassimatis wrote: And a TSM DB Backup takes (at least) one volume, so with physical cartridges, that's a whole tape. With VV's, you're only using the actual capacity of the backup, which is more efficient on space. At the cost of some reliability. What happens if the particular tape the virtual volumes are on goes bad, and you're in a disaster needing a DB restore? I'd rather spend the extra money on tapes and know that if something goes bad, we'll at least be able to recover some of the data ... (I'm seeing installations getting over 2TB on the new 3592/TS1120 drives - for a 60GB TSM DB Backup, that's VERY wasteful). Well, why not do what we're doing soon? We currently have some 1200 LTO2 tapes, and are in the process of migrating from LTO2 to LTO4; (some of) the LTO2 tapes will be kept in the silo for database backups (along with a single LTO2 drive for writing to those tapes.) There's another silo with LTO3 volumes; some of the LTO2 tapes will be put into that silo for exactly the same reason (LTO3 drives will write LTO2 tapes, so there's no issue with needing an LTO2 drive in that silo, at least for the time being). Call me a conservative fuddy duddy if you want, but I prefer to keep the TSM database backups as simple as possible.
Re: TDP for SQL question
On 03/08/2007, at 2:53 PM, Paul Dudley wrote: I have been told that if I want to create an archive backup of an SQL database via TDP for SQL, then I should create a separate node name in TSM (such as SQL_Archive) and then backup using that node name once a month (for example) and make sure to bind those backups to a management class that has the long term settings that meet our requirements. What else is involved in setting this up? On the client do I have to create another dsm.opt file with the new node name to match what I set up on the TSM server? In the context of the BA client, you can create multiple server stanzas in dsm.sys to point at different TSM instances and/or node names - eg, we do this with one particular system so that we can hold an image backup for ten days normally, but once a week, put one aside for eight weeks. The dsm.sys for this node looks like: servername (instance)_8WR commmethod tcpip tcpport(port) tcpserveraddress (server) passwordaccess generate [...] nodename (client)_8wr servername (instance) commmethod tcpip tcpport(port) tcpserveraddress (server) passwordaccess generate [...] nodename (client) All that is in dsm.opt is a line servername (instance), so that backups by default go to the client's standard node; for the 8 week retention, we add a -se=(instance)_8WR, and TSM does the right thing. (The '[...]' signals where I've cut out a number of TCP/IP options, along with other client-side options.) How to do this for a TDP client is left as an exercise for the reader. :-) Hopefully, it should give you some ideas. Good luck.
Re: TSM performance very poor, Recovery log is being pinned
On 29/07/2007, at 10:03 PM, Stapleton, Mark wrote: From: ADSM: Dist Stor Manager on behalf of Craig Ross TSM is installed on Solaris 10 This is something that popped right out for me. Do you have your storage pools located on raw logical volumes or mounted filesystems? If the latter, that might be your problem. Solaris has traditionally had incredibly poor throughput performance on mounted filesystems. You might give thought to rebuilding those storage pools on raw logical volumes. Of course, that will require that you completely flush all data from your disk storage pools to tape storage pools first, so as not to lose client data. A small trap for young players: TSM has constraints in place to stop it writing to cylinder 0 of a raw volume on Solaris. If you direct TSM at slice 2, or some other slice that includes cylinder 0, it will barf, and the error message is rather cryptic (sorry for the vagueness; it's been a year or so since I bumped into this. At the time, I was working on the storage pool volume level, but I would expect to see similar behaviour for a DB or log volume.) Workaround is simple: make slice 0 include the entire disk starting from cylinder 1, and use slice 0 as the raw volume. I am not going to enter into a debate about the relative merits of raw volumes versus files on filesystems, as I have insufficient direct knowledge to judge either way (I'm trusting a more senior colleague to make the right call there. :)
Re: Sharing a 3583 library
On 27/07/2007, at 1:43 AM, Zoltan Forray/AC/VCU wrote: Thanks for the reply. I already have things configured like this. What I was hoping for was zOS/MVS sysplex sharing smarts (for you mainframe folks). With a properly configured sysplex, the drives are configured and online to all systems at the same time and the OS figures out which drive is in use by which system and which drives are available and choses acordingly, not stepping on any other system. I was hoping the SAN type libraries were smarter and that multiple TSM library managers could just figure out what drives are available and use them accordingly. I wanted to avoid an all-or-nothing reconfiguration since I need to move library management/ownership to my new TSM server (phasing out the old AIX server currently owning the 3583's) and having to reconfigure all TSM servers that currently share the libraries. So if I understand you correctly, you will eventually be moving the library management completely to the new server, but you want to avoid reconfiguring all the existing TSM instances? I've just completed moving two library managers from one TSM instance to another (the new manager is dedicated solely to library and configuration management, where the old managers also served backup duties.) It turned out to be remarkably easy: * Delete the old paths, drive definitions and library definition on the old library manager (and the new library manager if it's currently a library client). * Define the library, drives, and paths on the new library manager (setting the drives to offline, so no tapes are accessed until you're finished.) * Checkin the libvols on the new library manager (CHECKIN LIBVOL X search=yes stat=scratch checkl=barcode) * Update the old library clients (UPDATE LIBRARY X PRIM=new_lib_manager on all instances) * Create a library definition on the old library managed (of type shared, pointing at the new library manager.) * Run an AUDIT LIBRARY on the library clients (including the old library manager). * Set the drives to online, and you're away. The audit on each client will set tapes with data to private status, owned by the relevant client. If you're paranoid about this, check them in as private, owned by the new library manager, and make a note of the scratch volumes known by the old library manager before deleting the library definition (query libvol). The audit will update the ownership to the correct node, you can manually update the remaining volumes to be scratch, and chase up the remaining volumes owned by the new library manager (assuming it didn't own any volumes beforehand) as potentially orphaned - we found some 30 tapes that should have been returned to scratch by the clients, but the library manager hadn't updated them for some reason so they were still marked as being private. As for the 6/8 character volume label: I can't speak for a 3583, but we're using a pair of 3584 libraries. In the web-based management system, there's a section for Library, Logical Libraries; have a look at the modify volser reporting option - it might help. Note that this will likely affect *all* volumes, so if you have data stored on volumes with a mix of 8/6 volume serial numbers, you're in trouble ... Hope this helps.
Re: TSM vs. Legato Networker Comparison
On 26/07/2007, at 2:54 AM, Schneider, John wrote: Greetings, We have been a TSM shop for many years, but EMC came to our management with a proposal to replace our TSM licenses with Legato Networker, at a better price than what we are paying for TSM today. This came right on the heels of paying our large TSM license bill, and so it got management's attention. We have an infrastructure of 15 TSM servers and about 1000 clients, so this would be a large and painful migration. It would also require a great deal of new hardware and consultant costs during the migration, which would detract from the cost savings. So instead of jumping from one backup product to another based on price alone, we have been asked to do an evaluation between the two products. Do any of you have any feature comparisons between the two products that would give me a head start? Funny you say this. Monash was a Legato (now owned by EMC) shop until the TSM migration around 3-4 years ago; I suspect that a large number of the problems that were perceived to be Networker's fault were actually the fault of the aging DLT silos and drives that underlay Networker. I still have fond memories of those silos; they gave me a great deal of callout pay every time they had a stuck cartridge or similar. :-) I also suspect that the greater reliability we've had since putting in TSM is more because we also got in new tape silos (LTO2, now half LTO2 and half LTO3, and soon to be half LTO3 and half LTO4) - if we'd stuck with the DLT silos, we'd still be in a world of pain, regardless of the software. There are plusses and minuses to both products. Some points to consider: * Networker uses the traditional full plus incrementals, or dump levels system. Monash used a pattern of full once a month; incremental every other day; and a dump level interwoven - so, for example, it might go full, incremental, level 8, incremental, level 7, incremental, level 9, level 2, incremental, level 8, incremental, etc. - the idea being to minimise the number of backups needed to restore a system. * Networker indexes are somewhat analogous to the TSM database. In theory, you can scan each tape to rebuild the indexes if they're lost; in practice, if you lose the indexes, you're pretty much dead - there's just too much data to scan if the system is more than moderately sized. Yes, Networker backs up the indexes each day. :) * At least the versions of Networker (up to 7.x) we used doesn't support the idea of staging to disk - everything goes directly to tape. However, data streams from multiple clients are multiplexed onto tape to get the write speeds up. This is good for backups, but does make recovery slower (since the data read will include a lot of data for other clients.) * No more reclamation or copy pools to deal with (because of the traditional full/incremental/dump level system). So the burden placed on the tape drives is probably going to be significantly lower (although you will be backing up more data each night than you would with TSM.) * I don't think Networker has anything analogous to TSM's scratch pool: volumes belong to a pool of tapes, and there's no shuffling between the pool. So if the standard pool has a hundred tapes available for use, but the database pool is out of tapes and needs one more, you need to manually intervene. This *may* have been because of the way we configured Networker, though, and it may also have changed in the interim. Note that you *have* to have a separate pool of tapes for index backups. My honest assessment mirrors that of the other people who have replied: use this as an opportunity to negotiate better pricing from IBM, and point out to the powers that be that there are risks involved with moving to a different backup product. There's nothing wrong with Networker, it's a good system, but you aren't familiar with it; it takes time with any new product to learn the tricks of the trade. It's only in the past year or two that we've started to feel more competent with TSM, as we've found and dealt with problems in the production system which never showed up (and would never show up) in the smaller scale proof of concept. You also should note that it took Monash a couple of years to finish the migration from Networker to TSM; I would expect a migration in the other direction would take at least a year. I definitely would not advise a dramatic cut-over - do a small number of servers at a time to make sure you're not pushing the server too hard (and besides, you want to stagger the full backups so they don't all take place on the same day ...) Oh, one other point that comes directly from Monash's experience with Networker (assuming you do go down that path): we had a number of large servers (mail in particular) that would take a very long time to do a complete full backup. We ended up setting Networker up to stagger the full backups on their filesystems: system filesystems on day 1; mailbox
Dealing with defunct filespaces.
Hi all. Whilst investigating something else, we discovered a number of nodes that have old filespaces still stored within TSM - eg: Node Name: (node name) Filespace Name: /data Hexadecimal Filespace Name: FSID: 4 Platform: SUN SOLARIS Filespace Type: UFS Is Filespace Unicode?: No Capacity (MB): 129,733.3 Pct Util: 92.1 Last Backup Start Date/Time: 06/09/05 20:03:56 Days Since Last Backup Started: 764 Last Backup Completion Date/Time: 06/09/05 20:05:16 Days Since Last Backup Completed: 764 Last Full NAS Image Backup Completion Date/Time: Days Since Last Full NAS Image Backup Completed: Node Name: (node name) Filespace Name: /Z/oracle Hexadecimal Filespace Name: FSID: 12 Platform: SUN SOLARIS Filespace Type: UFS Is Filespace Unicode?: No Capacity (MB): 119,642.2 Pct Util: 31.5 Last Backup Start Date/Time: 08/26/05 01:03:08 Days Since Last Backup Started: 686 Last Backup Completion Date/Time: 08/26/05 01:14:01 Days Since Last Backup Completed: 686 Last Full NAS Image Backup Completion Date/Time: Days Since Last Full NAS Image Backup Completed: Node Name: (node name) Filespace Name: /mnt Hexadecimal Filespace Name: FSID: 15 Platform: SUN SOLARIS Filespace Type: UFS Is Filespace Unicode?: No Capacity (MB): 120,992.9 Pct Util: 55.8 Last Backup Start Date/Time: 01/26/06 20:05:15 Days Since Last Backup Started: 533 Last Backup Completion Date/Time: 01/26/06 20:06:34 Days Since Last Backup Completed: 533 Last Full NAS Image Backup Completion Date/Time: Days Since Last Full NAS Image Backup Completed: These are all filesystems which existed at some time in the past, but which were removed as part of an application upgrade (or system rebuild, or ...), and hence no longer exist. It seems that TSM is taking the attitude of if I can't see the filesystem, I'll not do anything about marking files in that filesystem inactive, so the data never expires. I can understand the reasoning behind this approach, but it does mean that there's a large amount of data floating around that is no longer needed (a quick and dirty estimate says around 83 TB across primary and copy pools, although some of that needs to stay). A delete filespace will clear them up quickly, obviously, but there's a twist: how can we identify filesystems like this, short of going around to each client node and doing a df or equivalent? Searching the filespaces table gives us some 600 filespaces all up; I *know* that several of these have to stay - eg, image backups don't update the backup_end timestamp, and there are some filespaces that are backed up exclusively with image backups. At the moment, the best I can come up with is to: * use a SELECT statement on the filespaces table to get a first cut (select node_name, filespace_name, filespace_id from filespaces where backup_end current_timestamp - N days); * use QUERY OCCUPANCY on each of the filespaces mentioned in the first cut; if the total occupied space is below some threshold, ignore it as not being worth the effort; * use a SELECT statement on the backups table to confirm that no backups have come through in the past N days. (select 1 from db where exists (select object_id from backups where node_name=whatever and filespace_id=whatever and state=ACTIVE_VERSION and current_timestamp backup_date+90 days) -- I use exists to try to minimise the effort TSM needs to put into the query; I also have the active_version check in there for the same reason (if there's only inactive versions, they'll drop off the radar anyway in due course). Hopefully TSM's SQL execution is optimised to stop in this case when it finds one match rather than trying to find all matches ...) Does anybody have any better ideas? Unfortunately, because of the nature of Monash's organisation, simply having central policies saying you must do X when shuffling filesystems around won't cut it (and let's be honest
Re: Mixing drive types in logical library 3584
On 31/05/2007, at 3:38 AM, Chris McKay wrote: Hi all, I have been told by IBM that fibre LTO2 drives are no longer available. We wish to expand the number of drives in our 3584 library from 2 to 5. The current 2 drives are fibre LTO2 drives, would it pose a problem by adding an additional 3 LTO3 drives to that logical library? Is it possible under Windows?? I realize I will need to continue to use LTO2 media, as the existing LTO2 drives would not be compatable with LTO3 media. Our TSM server is running on Windows 2003. Compatibility with LTO media is straightforward: one generation back for write; two generations back for read. So an LTO3 drive is a perfectly reasonable way to extend your LTO2 capabilities, as well as giving you the ability to migrate to LTO3 down the road. At Monash, we're running one silo on LTO3, one on LTO2. We're looking at upgrading the LTO2 silo to LTO4, and have checked the compatibility; our information is that TSM is smart enough to mount LTO media only in drives appropriate to the operation (so LTO2 media will only mount in an LTO4 drive if it's being read, not written; LTO4 media will never be mounted in an LTO2 drive; etc.) So if you were to add LTO3 media to that library down the road, TSM would access it only through the LTO3 drives (and if all the LTO3 drives were being used with LTO2 media while an LTO2 drive sits idle, too bad.) You may want to double check this, though, if it's a concern for you down the road. I can't speak for how far this advice extends into Windows - we're running TSM on Solaris - but it's a starting point. My expectation would be that there would be no problem as long as the LTO3 drive is compatible with the drivers on the Windows system.
Re: Continuity of Data Backup
On 18/05/2007, at 5:01 AM, Avy Wong wrote: Hello, We have two instances TSM_A and TSM_B, Currently we want to move some nodes from TSM_A to TSM_B. Once the nodes are moved to TSM_B, the filespaces will be backed up over TSM_B. Is there a way to keep continuity of the backed up data ? Do I need to move the backed up data of those nodes from TSM_A to TSM_B? How do I go about it? Thank you. Easiest way to go about this would be to do an EXPORT NODE between instances, assuming a high speed link between the two instances. If the link between the two instances is slow (meaning less than the streaming read rate of the tape drives, or if the importing server is unable to write as fast as the tape drives can read), you'd probably be better off doing the export to a file storage pool, copying the exported files over to the other server, and importing them (tedious in the extreme, and needs a large chunk of temporary disk space, but ... well ...) Alternatively, if you don't have any important archived data for the nodes in question, you can just cut them over to the new instance, take the hit of the one-time full increment, and remove them from the original instance at some appropriate time, according to the retention policies of the company. This assumes you have enough tape storage space to take the short-term hit. Cheers.
Moving the TSM configuration manager
A bit of background: to keep the database size down, we have several TSM instances. One of these instances acts as a configuration manager; the others are all configuration clients (so to speak). For various reasons, we intend to create a new TSM instance dedicated solely to library management and configuration management - no TSM clients will backup directly to that instance. The major stumbling block I'm having right now is with policy domains; such a change means: * Copying policy domains to temporary PDs. * Moving clients from their current PDs to the temporary PDs. * Re-directing the configuration subscription to the new TSM instance. * Moving clients from the temporary PDs back to the permanent PDs. I'm aware of the implications this has for scheduling, and am not too concerned about it (just recreate the schedules, and all will be happy). I'm more concerned about data retention - given that the policy domains will have the same management classes, etc., after the cutover, will it affect retention in any way? I'm especially concerned about archived data; backup data will rebind as desired with the next backup if there's any difference, so I doubt that will be an issue. The documentation is not as clear on this matter as I would like. Any and all assistance would be gratefully received. Thanks, Stuart.
Re: More on fssnap on Solaris
On 11/08/2005, at 6:21 AM, Mark D. Rodriguez wrote: Stuart, I am not a Solaris expert so please bare with me on my suggestions. First of all you might try backing the filesystem up by using a virtualmountpoint, this option has worked for me in the past on unsupported file systems in Linux. Now in regards to your image backups. Again, I am more of a Linux guy then Solaris, but in Linux there is a mknod command the lets you create block or character special files, i.e. devices. Does Solaris have a similar command? If so why not just create unique device names that have the same device major and minor numbers as the fssnap devices. They would be effectively the same device! Then you could backup using those unique names and everything would then be kept straight. It would require some scripting but it could be done. I do hear where you're coming from, and yes, in theory it should be doable ... the main hassle, though, is that this would mean a fair amount of fiddling. I'd prefer to keep things as simple as possible, on the basis that that way, there's less to go wrong. Virtual mount points aren't really what I'm after -- yes, fssnap will probably let me do an incremental backup of a point in time, but I want to do an image backup, not a file-level backup. I do apologise if my comments weren't clear on that point. So whilst your comments are good, they don't quite get to where I want to go. Thanks for the suggestions nonetheless; they might well be of use in other applications. Just not in this one, alas.
More on fssnap on Solaris
I've done some experimentation. So far, I've found that TSM will not backup a filesystem image for a fs mounted from /dev/fssnap/N, but it will backup the image as a raw image. This then raises the question of consistency -- making sure that a given filesystem is backed up with the same name every time. As an example, suppose I have three filesystems (A, B, and C) that I want to backup as an image using fssnap. If I create a snapshot for A, it will be given the snapshot device /dev/fssnap/0. If I then back it up and delete the snapshot, and then create another snapshot for filesystem B, filesystem B will also get the fssnap device /dev/ fssnap/0. Conversely, if I create a snapshot for filesystems A and B concurrently, they will be given different numbers. (0 and 1). Subsequent creations of snapshots will give them the same numbers as they were given in the first instance (so if filesystem A has no snapshot, and I ask for B to have a snapshot created, it will be given /dev/fssnap/1 -- not 0.) I have not experimented beyond two simultaneous filesystems as yet. Based upon my experimentation and reading of the manuals, I get the feeling that there is no way to tell TSM, Back up this device, but name it as /foo/bar rather than /baz/bam. Can anybody shed any light that may be of use in backing up filesystems in this manner, whilst keeping it clear enough that we can easily determine which device is associated with a given filesystem, and that this mapping is kept consistent? Am I wrong? If not, I fear that this technique will be rendered useless to us. There are other options that I can pursue, as I've indicated before, and I'm about to start on them; I was hoping that the collected TSM gurus might be able to point me in another direction that I haven't yet noticed. The other option, of course, is to ask IBM to support fssnap in a future release, but since that's not likely to happen before 5.4 (at a guess) ... Thanks for all advice that you can give, Stuart.
Using fssnap for image backups on Solaris?
Just wondering: there are a couple of instances within the university where a system has large filesystems with a large number of small files. (eg: our mail spools have a single file for each email message ...) This is, obviously, a worst-case scenario for both backups and restores. I notice in the documentation for the 5.1 TSM client that snapshots are supported on Linux-x86, with the Linux Logical Volume Manager; unfortunately, these systems are running Solaris, not Linux. In short: fssnap creates a new device entry, which it prints to stdout, named /dev/fssnap/N (N being a number) for the block device (raw device is /dev/rfssnap/N). Reads from this device are redirected to a combination of the original filesystem and a backing store (which is where the old data is captured before being overwritten; it is defined during the execution of fssnap). Backups should then proceed from this device image, rather than the original volume. I'm guessing that this image is a consistent filesystem based at the time of the fssnap command (even if not, it would be preferable to backing up a raw, mounted, filesystem.). The ideal solution would be to use TSM's image backup facility in conjunction with fssnap. If this is not possible, an alternative would be to forcibly backup that device (with all the headaches that likely would then ensue should the device name change between invocations of fssnap). I'm not sure at this stage whether we would be using image backups in conjunction with standard incrementals, but I suspect that we will in most cases; there's at least one case where this will not be useful, though. So the questions: * Does anybody have any experience with this type of backup, and have any suggestions regarding pitfalls, etc.? * Is there any likelihood of TSM supporting fssnap for image backups in the near future? * Any general advice from those who haven't dealt with this specific scenario before, but who do have some idea of the problems we're likely to encounter with image backups? * Comments relating to the possibility that the raw device fssnap provides may differ from invocation to invocation? * Suggestions for alternative backup methods? We're also looking at using the FastT900's snapshot capability, which I imagine would have similar problems in large to using fssnap. Many thanks for any and all tips and pointers. A quick Google doesn't bring up much, and none of it seems to be useful. :(
Archives missing from archive list.
Greetings all. This one has me baffled, and whilst I'd normally spend more time investigating, I'm going on leave tomorrow for three days, and other parties view this as sufficiently urgent that I want to resolve it if possible before then. Basically, we have a number of applications based around Oracle databases (eg: SAP); these are backed up using the archive facility of TSM (yes, I'm aware of the inherent contradiction in that statement.) As an example of the problem, I have logs from the client that claim that the command: dsmc archive -description=CDUT_WEEKLY_07/05/05 01:00 - compression=no -archmc=DDB_WEEKLY_8WR_1 /space/CDUT/data/sis_d4.dbf completed successfully. Investigating the activity log for the appropriate time also indicates that the archive completed successfully. However, if I then run, on the archiving client, the command: dsmc q archive / -subdir=yes | grep CDUT_WEEKLY_07 I get back nothing -- there are entries for March or so, but definitely not for May. In comparison, if I run the command select description from archive where node_name='(source host)' and look through the output, I see CDUT_WEEKLY_07/05/05 01:00 in the list. All of this combined suggests to me that the data is there; it's just that the query archive command, for some unknown reason, isn't reporting it. Anybody have any suggestions on what to do next? Many thanks in advance, Stuart.
Re: Archives missing from archive list.
On 09/05/2005, at 1:34 PM, Ian Hobbs wrote: Make sure you are doing the q archive as the same userid (oracle perhaps) that performed the archive in the first place. Nice try, but I get back the exact same results: absolutely no mention of the (for instance) CDUT_WEEKLY_07/05/05 01:00 archive, even when the command is run as oracle. In fact, to the best of my knowledge, I get the same results as root as if I'm logged in (via su) as oracle. Unless using su to access the account in question makes a difference, but I would be *extremely* surprised if that were so. If that was all it was, I wouldn't have been prompted to look into this in the first place. :) My thanks nonetheless. [snippage]
Re: Archives missing from archive list.
On 09/05/2005, at 2:35 PM, Ian Hobbs wrote: This could be a really long shot. try dsmc q archive {insert the source filespace name here}/* - subdir=yes | grep CDUT_WEEKLY_07 It appears that the request is for any archive off the root filespace. Possible. I'm quite away from a unix system at the moment, and was only able to check using my home windows server. I've been getting varying results that don't make much sence. example, the dsmc q ar * -su=yes show me all archives from my C drive, but not my E drive, but if I do a q ar e:\* -su=yes then I see the E drive archives. H Bingo. Thank you *very* much for that. I'll pass that on to the DBA; hopefully that will calm her down. :-) It would be nice to know why this is happening, but in the absence of that knowledge, I'll take any solution I can grab. Especially since it seems to give some of the older archives from the same filesystem... *shrugs* Again, many, many thanks. Stuart.
Re: Export feasibility
On 11/03/2005, at 6:25 AM, Ragnar Sundblad wrote: --On den 10 mars 2005 16:18 +1100 Stuart Lamble [EMAIL PROTECTED] wrote: On 10/03/2005, at 8:15 AM, Ragnar Sundblad wrote: [snipping] I think that the most logical way to accomplish this with TSM is to do a complete export like export server filedata=allactive ... What's wrong with using backup sets? The end result is the same -- all currently active data is stored on tape and can be moved offsite, without needing any maintenance from the main TSM server. Backup sets may be fine too. There are two issues with it that made me think that they were not as suitable for this as an export: - There seem to be no way to generate a backup set that contains multiple nodes, so I take it that if I don't want one tape per node (hundreds), I will have to generate the backup sets to disk and find some other way to pack them on tapes. Am I wrong here? Not as far as I can tell, alas. The other is probably not as important -- I agree that symmetry is nice to have, but for DR purposes, you can probably live without it. [more snipping] I was just thinking that there might be some other resource that that server could get short of, like database rollback or something, if it was both doing an export for days and at the same time was backing up. I hope that is not a problem, and especially it shouldn't be if I use backup sets instead. On that, I'll have to defer to the more knowledgeable crowd on the mailing list; I honestly don't know. One option you do have is to export in batches -- say, a quarter of the nodes in one batch; another quarter in the next batch; and so on, giving the server time to do any maintenance it needs but can't perform whilst an export is occurring. I'd be surprised, though, if there was any of significance. Not saying there won't be any, just that it would surprise me. If backup sets (see the generate backupset command in the admin reference manual) fill your needs, my advice would be to make use of them, rather than using the rather fragile option of server data exports. I very much agree, simple is good! :-) It is interresting, but sad, to hear that you too find exports fragile. May I ask what the problems typically are? I've found, for example, that if the server is unable to read the data it wants off the primary tape, it doesn't appear to revert to the backup copy to complete the export; rather, the export fails. I haven't chased this up, as I had far too many other (rather urgent) matters on my hands when it occurred, so there may be other factors at play -- but that's one example. There are a couple of other things that may or may not be of concern; again, I haven't spent as much time as I should verifying all the ins and outs, so without knowing, I'd prefer to keep my mouth shut on them lest I be seen as a fool. :-) If I can do what I want with backup sets instead, that is probably a better way. One suggestion I do have would be to test. If you can afford to have the backup system in a potentially not-working state for a period of time, run some tests and see what happens. Try to make it break as much as you can; take tapes out of the silo (telling TSM about it, of course) to see how it copes; things like that. You'll end up being much more secure in what you're doing if you know how it can break, and how to recover if it does break. All this, of course, is with the understanding that, more likely than not, you'll not be in a position to do any of this. :-( Hope this is of help in some small way.
Re: Export feasibility
On 10/03/2005, at 8:15 AM, Ragnar Sundblad wrote: We don't want to give up storing away a complete snapshot of our systems off site every few months, over time maybe reusing the off site tapes so that we finally save a snapshot a year. I think that the most logical way to accomplish this with TSM is to do a complete export like export server filedata=allactive ... What's wrong with using backup sets? The end result is the same -- all currently active data is stored on tape and can be moved offsite, without needing any maintenance from the main TSM server. Given LTO 3 with an maximum data speed of 50 MB/s uncompressed, a moderate guess (I hope) for the data transfer rate would be 30 MB/s, which would give 93 hours to write it down. Given that a tape has a capacity of 400 GB uncompressed, it would take at the most 25 tapes. We would obviously want to be able to use the backup server as usual while doing the export. Then you would need to make sure that your server has enough capacity, in terms of tape drives (and connectivity to the tape drives), network bandwidth (to cover both the export and the backups), and such like to cope with both the export (or backup set generation) and regular backups simultaneously. From my point of view, being in the middle of fiddling around with exports (server to server in my case) for various last ditch DR systems, I'd suggest keeping it simple. If backup sets (see the generate backupset command in the admin reference manual) fill your needs, my advice would be to make use of them, rather than using the rather fragile option of server data exports.
Export node direct to managed server problems.
Hey guys. We have four TSM servers (long story, don't ask) across two sites, setup with one as a config manager, the other three all subscribing to the config pushed out by the manager. For various DR reasons, we're wanting to do data exports from those four servers to a fifth server, located at a completely different site to any of the four main servers. These exports are for a limited subset of the servers being backed up, and there's no easy way to separate them from the not-to-be-exported hosts in terms of policy domains, storage pools, etc. At the moment, the intent is to use the export node command to push the data across directly to the fifth TSM server. I've defined the server links and the requisite administrators. I've also -- by way of experimentation as much as anything else -- set up the fifth server to subscribe to the same config pushed to the other servers. I'm now finding that, whenever I try to export a server directly to this fifth server, I get a bunch of error messages in the logs. Some of them are relatively spurious[1]; others, however, are of concern[2]. I've had similar problems previously, which were rectified by changing the version of the TSM server software running on each host to match (they're all running version 5, release 2, level 3.2 now). I'm pretty sure these problems only started after I made the fifth server a managed server, rather than being mostly independent. So. Questions. * Can a server to which data is being exported directly over the network be managed, or must it be completely independent? * If it can be managed, how can the spurious errors about managed objects be suppressed? * If it can only be managed for certain config items, what are the restrictions on what can and cannot be managed? * If it _can't_ be managed, I'm curious to know why not. (This one doesn't really matter beyond satisfying my curiosity :) * Presumably, there is no problem with making the server link information for the importing server managed -- that the issue is purely with making the importing server managed by one of the exporting servers? Having managed profiles would make a number of things a lot easier, but if I have to live without them, so be it. Also, if there are any trips or traps that would be useful to know, I'd greatly appreciate knowing them. :) If there are any details that need to be disclosed for this one to be tracked down, let me know; I'm well aware that the above is rather vague, but it's hard to know when you're just setting out what might be relevant and what might not be. Thanks for any and all advice you can give, Stuart. [1] For example: 01/25/2005 11:14:45 ANR0646I IMPORT (from Server TSM1): DEFINE CLIENTOPT: AMCommand cannot be executed - option set NW is a managed object. (PROCESS: 27) 01/25/2005 11:14:45 ANR0730E IMPORT (from Server TSM1): Internal error from AMcommand 'DEFINE CLIENTOPT NW INCLEXCL exclude sys:\vol$log.err FORCE=Yes SEQNUMBER=11'. (PROCESS: 27) 01/25/2005 11:14:45 ANR3229E DEFINE CLIENTOPT: Command cannot be executed - AMoption set NW is a managed object. (SESSION: 4560, PROCESS: 27) [2] Prime case in point: 01/25/2005 11:14:45 ANR0730E IMPORT (from Server TSM1): Internal error from AMcommand 'REGISTER NODE SOMENODENAME DOMAIN=NW_PD CONTACT= COMPRESSION=CLIENT ARCHDELETE=YES BACKDELETE =NO URL=SOMENODENAME.its.monash.edu.au:2002 CLOPTSET= NW FILEAGGR=YES AUTOFSRENAME=NO VALIDATEPROTOCOL=NO DATAWRITEPATH=ANY DATAREADPATH=ANY SESSIONINIT=CLIENTORS ERVER TYPE=CLIENT'. (PROCESS: 27) 01/25/2005 11:14:45 ANR0728E IMPORT (from Server TSM1): Processing terminated AMabnormally - internal error. (PROCESS: 27) 01/25/2005 11:14:45 ANR0620I IMPORT (from Server TSM1): Copied 0 domain(s). AM(PROCESS: 27) 01/25/2005 11:14:45 ANR0621I IMPORT (from Server TSM1): Copied 0 policy sets. AM(PROCESS: 27) 01/25/2005 11:14:45 ANR0622I IMPORT (from Server TSM1): Copied 0 management AMclasses. (PROCESS: 27) 01/25/2005 11:14:45 ANR0623I IMPORT (from Server TSM1): Copied 0 copy groups. AM(PROCESS: 27) 01/25/2005 11:14:45 ANR0624I IMPORT (from Server TSM1): Copied 0 schedules. AM(PROCESS: 27) 01/25/2005 11:14:45 ANR0625I IMPORT (from Server TSM1): Copied 0 administrato AMrs. (PROCESS: 27) 01/25/2005 11:14:45 ANR0891I IMPORT (from Server TSM1): Copied 0 optionset AMdefinitions. (PROCESS: 27) 01/25/2005 11:14:45 ANR0626I IMPORT (from Server TSM1): Copied 0 node AMdefinitions. (PROCESS: 27) 01/25/2005 11:14:45 ANR0627I IMPORT (from Server TSM1): Copied 0 file spaces 0 AM
Re: Export node direct to managed server problems.
On 25/01/2005, at 11:31 AM, I wrote: I've defined the server links and the requisite administrators. I've also -- by way of experimentation as much as anything else -- set up the fifth server to subscribe to the same config pushed to the other servers. I'm now finding that, whenever I try to export a server directly to this fifth server, I get a bunch of error messages in the logs. Some of them are relatively spurious[1]; others, however, are of concern[2]. Does anybody know where I can find a supplier of soft brick walls? *sighs* Looks like the problem's solved: I forgot to activate the policy sets on the fifth server. Talk about your trivial fixes ... Please excuse me; I have some banging of heads (mostly my own) to do.
Selectively duplicating client data across servers
Hey ho. Here's the skinny. We will, eventually, have a number of clients backing up to a TSM server on a regular basis (we're still setting up the SAN and other ancillary things that are needed to support the TSM server). Some of them will be filesystem backups; others will be database backups (which, if I understand correctly, are most likely to be seen as archives rather than backups as such). The setup involves two sites, A and B, which have multiple gigabit fibre connections between them (so they're effectively on the same LAN; the only real difference is a very small amount of additional latency.) Systems at site A will backup to a server at site B, and vice versa. The project I'm working on involves a third site; call it site C. The connectivity from sites A and B to site C is significantly poorer than that between A and B, largely because site C is significantly remote to A and B (whilst A and B are within a few km of each other), precluding running of fibre to it. (Not sure what the current connections are; that's not my area.) The idea is that if things go boom in a major way, and we lose everything at _both_ site A and site B, we want to have copies of the university's critical data (student records, SAP, that sort of thing) available for restore. Maybe the data will be a little old, but better than nothing. So the idea is this. The servers holding the critical data backup to their backup server as normal. Once every so often (once a week, once a fortnight, once a month...), we want to take the data off the _backup server_, and copy it to another TSM server at site C. We want to do this only for those critical servers, not for all servers. The data may be in the form of either archives or filesystem backups, or some combination of the two. We _don't_ want to be running extra backups for the servers in question to provide this redundancy. The two solutions I've come up with involve data export, and copy pools (via virtual volumes). The problem is, both of those operate at the storage pool level; there's no way to specify copy/export only the data for _this_ client, and no others that I can see. It's preferable that we not have to create separate storage pools (all the way down to the tape level) for these systems just so we can do this -- we'd prefer to have one disk pool and one tape pool for the whole shebang if possible. Backup sets may be doable, but I'm a little uncertain about how they'd go with virtual volumes, and also whether they'd cover archive data sets. So the question is: is there any way we can say to TSM Copy this client's data (backup and archive) to that server over there, ignoring all other client data in that storage pool? Or am I smoking crack (again)? Any pointers or suggestions on where to look would be very gratefully received. For me, this isn't so much a learning curve as a learning cliff. :) If it's relevant, the clients in question are mostly Solaris, and we'll be running Tivoli Storage Manager 5.2 on Solaris servers. Many, many thanks for any tips. I'm coming from a background with Legato Networker, and I'm still trying to get my head around some of the more interesting aspects of TSM, learning as I go. :) Cheers, Stuart.
Re: Selectively duplicating client data across servers
(Once more, this time with the _right_ From address. Sigh.) On 26/08/2004, at 12:40 PM, Steven Pemberton wrote: On Thursday 26 August 2004 09:19, Stuart Lamble wrote: Hey ho. Here's the skinny. We will, eventually, have a number of clients backing up to a TSM server on a regular basis (we're still setting up the SAN and other ancillary things that are needed to support the TSM server). Some of them will be filesystem backups; others will be database backups (which, if I understand correctly, are most likely to be seen as archives rather than backups as such). The setup involves two sites, A and B, which have multiple gigabit fibre connections between them (so they're effectively on the same LAN; the only real difference is a very small amount of additional latency.) Systems at site A will backup to a server at site B, and vice versa. So, you have the following? 1/ Clients at A backup to TSM server at B. 2/ Clients at B backup to TSM server at A. Yup, pretty much. Where are you producing the copypool versions? Eg: 1/ Client A - TSM B (primary) - TSM A (copy) ? 2/ Client A - TSM B (primary) - TSM B (copy) ? 3/ What copy pools? :( Oh, option three of course -- we want to save money on media! *ducks* Seriously, though, it'll be option 2, is my understanding -- if site A goes down, and some of the media at site B is dead, we'd like to still be able to recover the data. If we lost two sets of media at the same time, well, we're obviously not meant to have that data any more. (Cue the story I heard whilst doing Legato Networker training: a site with several copies of key data. Key data goes poof. First copy on tape is bad. Second copy on tape is bad. They send for the offsite copy. Courier manages to have an accident, and the tapes are ruined...) (The scary thing is, there were a couple of people advocating no copy pools for some of the clients. Thank God _that_ got shot down in short order.) [verbose description of the basic plan snipped in favour of Steven's summary] Something like this? 1/ Client A - TSM B (primary) - TSM A (copy) (all) - TSM C (copy/export) (critical only) A healthy paranoia. :) That's pretty much it. A more accurate picture would be: Client A - TSM B (primary/copy) (all) - TSM C (copy/export) (critical) Client B - TSM A (primary/copy) (all) - TSM C (copy/export) (critical) And remember: just because I'm paranoid, it doesn't mean they're _not_ out to get me... ;) [copying data from the original backup server to the remote site] The two solutions I've come up with involve data export, and copy pools (via virtual volumes). The problem is, both of those operate at the storage pool level; there's no way to specify copy/export only the data for _this_ client, and no others that I can see. Actually, you can export node for individual hosts, but I'm not sure if it's the best way to do what you're planning. However, export node can specify a client, time range, backup and/or archive data, active files only, and export/import directly via a server to server connection. Hm. Going to have to re-read the manual on that; I must have missed that point. *flick flick flick* ... ok, I missed that point. Excuse me whilst I carefully extract my foot from my mouth. :) It's preferable that we not have to create separate storage pools (all the way down to the tape level) for these systems just so we can do this -- we'd prefer to have one disk pool and one tape pool for the whole shebang if possible. I'd normally recommed that you DO create multiple storage pools, so that you can better control the physical location of the backup data. This can improve recovery performance by separating critical clients to their own storage pools and tapes. With only one huge disk/tape storage pool hierarchy each client's data will tend to fragment across a large number of tapes (unless you use collocation, which may greatly reduce tape efficiency instead). Interesting point. Everybody here is an utter newbie when it comes to TSM; we've done the initial training course (you should remember; IIRC, you were the one taking the course I was on :) which is all fine and dandy, but it doesn't really expose you to the little tricks of the trade which come up when you're actually _using_ the product. :) (And besides -- after too many months of not using the product because of wrangling that's out of the hands of the techies, you tend to forget the finer points that were covered on the course.) Still, I have a fair amount of faith that TSM will do the job we need; it's more or less a matter of what problems we run into along the way (and don't tell me we won't run into problems -- we will; it's just a question of how severe they are and how difficult to fix. With luck, they'll be less than what we have with our current backup system.) We've already ruled out collocation for the most part; I seem to recall an upcoming version of TSM has a weaker form of collocation (along the lines of group