home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!elroy.jpl.nasa.gov!usc!rpi!uwm.edu!ogicse!usenet.coe.montana.edu!osyjm
- From: osyjm@cs.montana.edu (Jaye Mathisen)
- Newsgroups: comp.unix.ultrix
- Subject: Re: DECnsr, has anyone any experience with it?
- Message-ID: <1992Aug19.224712.27925@coe.montana.edu>
- Date: 19 Aug 92 22:47:12 GMT
- Article-I.D.: coe.1992Aug19.224712.27925
- References: <JOHN.92Aug18105157@sekrit.WPI.EDU>
- Sender: usenet@coe.montana.edu (USENET News System)
- Organization: CS
- Lines: 174
-
- In article <JOHN.92Aug18105157@sekrit.WPI.EDU> john@sekrit.WPI.EDU (John Stoffel) writes:
- >
- >Our site is looking into getting DECnrs, the Network Save and Restore
- >system for our Network of DS3100s and 5000s. Has anyone out there had
- >any experience using this product? If so, could you answer some or
- >all of the following questions? Any other Backup and Restore programs
- >or philosophies or even anecdotes would be very appreciated. Thanks!
-
- Be happy to... (Be warned, critical commentary on a DEC product ahead).
-
-
- >What problems did you have during installation and configuration?
-
- None. It installed easily. Configuration is easy if you ignore the
- idiotic curses interface, and just use 'edit' from the nsradmin> prompt.
- You can block copy stuff right and left (remember to remove the
- resource ID gunk), and basically whip it out.
-
- One note, after you delete a resource, save it, let NSR update its databases,
- then re-edit and make more changes. If you don't you can get into situations
- that have inter-dependencies, where your only recourse is to exit w/o saving
- your additions. Of course, careful planning will help...
- >
- >What backup medium are you using?
-
- 2 CCG 8mm's (8200's), probably switching to 2 TLZ06's.
-
- >How much disk space are you backing up, and how much space did the
- >online file indexes use?
-
- 12-14 GB's. The indexes are a big problem... more later.
-
- >Was it easy to use? Easy to add/delete partitions from the list of
- >disks to backup? Easy for users to ask for old files to be restored?
-
- all things considered, yes. yes. yes.
-
-
-
- The biggest problem is the file indices. You can get into situations where
- the file index is so large, that NSR can't scavenge records from the old
- database, because their isn't enough disk space.
-
- I had 2 weeks worth of backups, 20 tapes, 6-7 clients, and it consumed
- all of an RZ25. (about 3.5 MB's left). Now, I wanted to nsrck the indices,
- but 3.5 MB's isn't enough space, so the admin has to clear out 500 MB's of
- disk somewhere else, copy the old db to the new space, make a symlink to
- point to the new database, run the nsrck, then copy it all back.
-
- I want to keep 3 months worth of indices and tapes on-line. That'll be
- about 1.3-1.4 GB's. Ultrix can't handle stripe sets or any disk large than
- 2 GB's, so I can't rebuild this index file if I need to. Major boner.
-
- Get the fastest disks you can to keep the index files on. I deleted a volume
- from the media database, and purged all the file index entries for that
- tape, it took 2.5 HOURS... on an RZ25. If you ever wanted to purge a bunch
- of tapes, you'd be better off to delete nsr, reinstall it, and rescan the
- tapes you want to keep with scanner.
-
- A nsrck of a 240 MB index file took 11 hours on an RZ56, on a machine that
- was otherwise unloaded. This seems extraordinary.
-
- Bitches:
-
- Doesn't support HP9000/700 series. Even after 2 releases. They support
- 300's and 400's, but who cares about them? How hard can it be to re-compile?
- I think DEC is chicken... don't give 'em any reason to buy the competitor...
-
- Big bitch and moan:
- Licensing is bizarre. My license expired for nsr, and now the nsr daemon
- won't start... Understandable... How the hell am I supposed to recover
- files 2 years from now when I decide to use HP Omniback and not NSR? I have
- to keep an active license for a product I won't use?
-
- It may be possible to use scanner and the uasm modules to restore from the
- expired tapes, I haven't had a chance to try it, but even if it works, it's
- beside the point. I can't use recover to get the data back, and browse the
- tapes looking for files, ... Seems to me a more reasonable restriction would
- be that w/o a valid license, saves won't start, but recovers will work. This
- is probably what will keep me from purchasing it again. I'll use amanda before
- this...
-
- When using mm to delete a volume, the mm command eventually returns, giving
- you reason to believe that the volume info has been purged, but it's not,
- it's still running in the background. But you don't know this, you do
- another mm -d, and now you've got two of them, and your nsr disk is going
- bonkers, ... it's a nightmare. To be fair, the "extra" nsrindexd's are
- just hanging around waiting, they run in sequence, not in parallel...
-
- No easy way to tell which file systems are backed up at which times, w/o
- following through a bunch of junk, ie, first, find the client that has
- the file system you want, then find which group it's in, then find that
- group specification, then find the time. Pain. It might be possible to
- give a query from the nsradmin prompt that would do this for you, but it
- didn't look trivial. A simple tool that would generate a calendar for a month
- or week that has the schedule of clients/groups for quick perusal would
- be nice.
-
- Requires that the server machine be in the clients /.rhosts to start up
- remote saves. I had hoped for a more sophisticated mechanism... Seems
- like it's opening up the client w/o a real good reason... why not incorporate
- some new authorization technology into this thing?
-
- It's based on Sun RPC, rather than OSF technologies... Perhaps DEC is
- only paying lip service to incorporating OSF DCE technology...
-
- NSR provides a mechanism for specifying how many parallel saves will be
- done at once. Don't believe it. It says 4 is the default, and it looks like
- 4, but I've seen 8 at a time all trying to run. Buries the server.
-
- NSR's got memory leaks, and it's a major memory hog... my daemon reached 13MB's
- of real RAM in size. The server is a 5000/240 w 48MB's of RAM, and during
- backups, the free RAM drops to 2-3 MB's.
-
- I did not get a chance to test 1.1, perhaps some of these problems are fixed.
-
- No Motif interface... It would be nice to be able to just point and click
- your way through, but doesn't happen.
-
- The monitoring software is based on curses, and gets easily confused about
- what to display. It has a screen that's divided into several areas, and
- has it's own ideas about how they should be sized. You have to make the
- window pretty large to get all 8 sessions that might be running at once to
- be displayed, but it grows the "message" area way out of proportion. Then
- it prints 6 copies of the status messages to the message area, scrolling
- the useful info off too fast anyway. (Not all messages, just (Info) messages
- seem to get multiple copies, and fast). The curses interface also gets
- confused, it will show a backup in progress in the "current sessions"
- area, but won't show it in the message area, and if you resize the window,
- then the message will go away.
-
- NSR doesn't support many devices... None of the SCSI TF devices, no
- 8mm 8500's, no TLZ06 support in compression mode. NSR can't stream
- both drives at once, it fills one, then starts the other.
-
- You can't specify where you want file systems to go, ie, Full backups
- on one tape devices, incrementals on the other.
-
- It's tough to make off-site storage tapes, ie, I want to make a tape
- that has all the user files systems backed up, and gets put in a vault, but
- doesn't become a part of the regular backup indices. You can't do this.
- What you have to do is empty and dismount your tape drives. Disable all
- the groups so that there's no chance of an automatic backupstarting, then
- init and label a new tape, do your backup, the delete the volume and
- file index entries from the media database, renable the groups, load up
- the old tapes, and check that you didn't miss any scheduled backups.
- Pain in the butt.
-
- Over all, I'd give it a grade of B-, it's better than dump/restore, but
- has some major pain-in-the-butt areas that need polish.
-
- Would I buy it again? Doubtful. I've chatted with people who have some
- of the other enterprise backup solutions, and they don't report having
- any experiences like this.
-
- Minor side note to EDU people: NSR isn't under CSLG/TEI stuff yet, although
- it's a dec-developed product. It's not outrageously expensive, but it
- ain't cheap either. With my luck, it'll go under CSLG 2 days after I buy
- something else.
-
- Am I too critical? Too picky? I don't know. Some of these problems seem
- very obvious... Rumor has it that DEC's been using NSR internally
- for quite some time. I hope their license never expires :)...
-
- To DEC's credit, they've been helpful in answering questions, even if
- most of the answers have been, "you can't do that."
-
- If you have questions or commentary, feel free to mail...
-
- (NSR is kind of like DEC's VXT2000 X terminals. So darn close, yet so far)...
- --
- Jaye Mathisen, COE Systems Manager (406) 994-4780
- 410 Roberts Hall,Dept. of Computer Science
- Montana State University,Bozeman MT 59717 osyjm@cs.montana.edu
-