toasters

toasters@lists.teaparty.net

13515 discussions

Re: reboot of TLU -> reboot of fileserver
by sirbruce＠ix.netcom.com 23 Apr '98

23 Apr '98

On 04/23/98 08:31:55 you wrote: > >Hi, > >I wanted to check out the following issue that was brought up in >one of the responses to my poll (still waiting for more replies): > >When a TLU (tape Library Unit) is directly attached to a fileserver, >does every reboot of the TLU necesarily mean you must reboot the fileserver? No, it doesn't. At least with the TLUs I've used (Quantum DLT-4700, Breece Hill) a reboot of the filer was not necessary if the TLU was power-cycled. However, there was rarely any need the power cycle the TLU in the first place. Note that you can't properly "hot swap" a tape device on an off a filer, but I have seen it work before, although that's not properly supported and probably depends a lot of the specific filer, tape device, OS, and the phase of the moon. :) > >OR > >Is this only specific to the TLU vendor ( as suggested below). I think it's specific to that TLU vendor. >I see this as a serious issue!! >I don't want the uptime of my fileserver to be affected >by something like this. Understandable. I don't think it's different from any other backup server in this regard. >| You mentioned a very interesting note - attaching a jukebox >| to the fileserver might mean that you will need to reboot the file >| server from time to time. WHY? >| When you reboot the tape library juke box - does this force >| you to reboot the file server as well? > >If the stacker/jukebox hangs for whatever reason (robotic arm, etc), >and you need to power-cycle the tape device, you cannot do so with >the filer on. You run the risk of frying the scsi bus. I find this difficult to believe, considering the SCSI bus has to deal with, for example, the power supply on shelf suddenly failing, or the power supply of the stacker itself, etc. If one considers the case of a short power outage, you can't control wheter the filer or the stacker will get power first when it comes back on, so sometimes the filer will be on first. Surely equipment is robust enough to support this. >Now, it could be that the problems I had were related to the specific >DLT stacker (an ADIC vls library) we used. I don't know. I have not >risked that sort of problem any more. The risk, if there is one, would have to be very small. No more than the risk for any other incident relating to power, and the number of SCSI busses fried in those instances must be small. If you have a stacker that is hanging frequently, you're much better off worrying about fixing the stacker's hanging than the small chance of a SCSI bus being fried if you power-cycle. See also my earlier comments about the filer possibly having to be rebooted if you can't abort a dump/restore that was in progress when the stacker hung. But that is a software issue, not a hardware problem. Bruce

1 0

Budtool to backup netapp
by Alan Phoenix 22 Apr '98

22 Apr '98

Has anyone tried using Budtool to backup a netapp? Al

2 1

subscribe
by David West 22 Apr '98

22 Apr '98

1 0

Re: Restoring A LOT from snapshot
by sirbruce＠ix.netcom.com 22 Apr '98

22 Apr '98

On 04/22/98 09:37:47 you wrote: > >In message <199804221400.KAA21557(a)prospero.dma.isg.mot.com>, Christoph Doerbeck > writes: >> Solution 1: use 'find' and 'cpio' and transfer all the data back. this >> would appear to require 2x current capacity of the filer to replicate >> all data out of snapshot. if you're using >50%, this method would >> fail. > >If the files in question have been removed, rather than modified, this >will only copy the files that have been removed. On other files, it >will either refuse to copy because it thinks its copying a file onto >itself. It will complain the whole while (slowing things down), but it >will only add the space consumed by the deleted files. At least that's >what happened when I tried it in my home directory. An interesting observation. One example where having the snapshot FSID the same as the filesystem FSID is actually an advantage. :) Bruce

1 0

Restoring A LOT from snapshot
by Christoph Doerbeck 22 Apr '98

22 Apr '98

Question: When restoring from snapshot, is there a method which will not recreate data in the real data area. Example: Random files are removed throughout the filer. We decide to backtrack to Mondays snapshot. We plan to copy everything out of snapshot back to the real data area. Solution 1: use 'find' and 'cpio' and transfer all the data back. this would appear to require 2x current capacity of the filer to replicate all data out of snapshot. if you're using >50%, this method would fail. Solution 2: use 'find' and 'diff' and 'cpio' to enhance solution 1 with a check. Now only files which differ from the snapshot will be replaced. Solution 3: dump the snapshot to tape, and then restore from tape after removing existing snapshots. Solution 4: boot from floppy, use prev_cp and wack? Has anyone had any experience with 'prev_cp' and 'wack'? We're leaning at #4. Any comments would greatly be appreciated! ######################################## # Christoph Doerbeck # Motorola ISG # email: doerbeck(a)dma.isg.mot.com

2 2

Re: backup times
by sirbruce＠ix.netcom.com 22 Apr '98

22 Apr '98

On 04/22/98 11:15:20 you wrote: >Because if you have 8 filers, most, potentially all with > 20GB of >compressed data, putting an autochanger on each filer is prohibitively >expensive. Going the other route and putting, say, a DLT7000 on each >filer is also nuts. Cost for one of those around here (toronto, >ontario, canada) is something like $12500 plus the cost of a tape >monkey to run around and change tapes daily. While I agree that ultimately network backup is a cheaper and easier to manage alternative solution, I think you overstate the case against local backup. There are several key points that you need to be aware of in a network backup solution: 1. If you have a good (read - frequent) backup scheme, and your filers are large, you won't save on the number of DLTs you buy. Consider the case where those 8 filers all have to be backed up once/day, and it takes about a day to run the backup. You'll have to have just as many DLTs streaming away on that backup host as you would if each one was attached to it's own filer. Now, a more realistic example would probably allow you to cut the number of DLTs in half, but you will have to make sure your schedule everything properly. And can recover quickly if you have a failure. 2. You probably can't afford to saturate whatever network you already have, so you'll need a dedicated backup network. So gigabit is overkill, but once you factor in your 100mbit hubs, the possible need to buy additional network interfaces on your filer and the backup hosts to support this, not to mention the backup host itself, the cost differential is less than it first appeared. 3. You SHOULD have a tape monkey running around to change tapes daily. You don't want those tapes sitting there; you want them labeled and catalogged and shipped off-site. This doesn't change if you have a central backup solution, and at the very least you'll still have to change tapes just as often. Unless you get a library with a bigger tape capacity for your central server than you would for an individial stacker for each filer, which again is a price consideration. 4. You will incur a performance penalty, although with the more recent releases that penalty may be less than 10%. So, IMHO, network backup is probably more expensive at 4 filers, cheaper but not worth the hassle at 6, and only becomes really preferable at, say, 8 filers. But these numbers all depend on how big your filers are and how frequently you want to do your backups. Bruce

1 0

backup and Restore questions
by Menachem Kaiser 22 Apr '98

22 Apr '98

Hi, We are currently evaluating a network appliance solution for our site. We still have deep concerns about our abilities to backup such vast quantities of data from one source and even greater concerns on our ability to restore such vast quantities of data in the event of a disaster within a feasable time frame. I would like to hear from people about how they backup their systems - i.e.: What backup software do you use? What size is your file server and what size are the partitions? Is the tape directly connected to the file server or do you backup via the network? If you backup via the network, how is your file server interfaced to the network (dedicated backup network(s)? - 10/100 Mb/s?) Do you use ndmp? How many streams of backup do you use in parallel? What speed/rate do you backup (GB/Hr) per drive? Do you utilize more than one tape drive for backups? Have you needed to restore a file system? If so, how long did it take you (GB/Hr) ? Were you able to retore in parallel streams? Since the list of questions came out long, I will be more than happy to call you and poll you verbally on the above (send me your number). I will even be more than happy to post the statistical results of those willing to participate.

5 4

Re: backup and Restore questions
by Mark D Simmons 22 Apr '98

22 Apr '98

> Hi, > > We are currently evaluating a network appliance solution > for our site. > > We still have deep concerns about our abilities to backup > such vast quantities of data from one source and even greater > concerns on our ability to restore such vast quantities > of data in the event of a disaster within a feasable time frame. > > I would like to hear from people about how they > backup their systems - i.e.: > > What backup software do you use? A cron job on an admin host kicks off a full-filer dump on each of our erm 9-going-on-10 filers using rsh and loggid the output to backups-logs. > > What size is your file server and what size are the partitions? Ranging from piddly 50GB ones to reasonably meaty 200MB. > > Is the tape directly connected to the file server > or do you backup via the network? We have 4 Exabyte stackers, 3 DLT4700's and a single DLT4000 all directly attached to filers. When there's a tape drive problem (infrequent but hey they involve lots of moving parts) we get a little downtime during lunch or beginning/end of business and perform a quick halt-detach-reboot cycle, work on the drive while providing service, then do a halt-attach-reboot to bring it back together. Sometimes this stalls a lot of machines and it'd be ideal if we didn't have to do it but pretty much everything recovers, though we have to get brutal with locks the filer wishes to honour after reboot when in actual fact the client has just gone awol and won't talk to the filer about the lock... > > If you backup via the network, how is your file server > interfaced to the network (dedicated backup network(s)? - 10/100 Mb/s?) We've considered the option but we'd need a dedicated LAN to get the performance and avoid loading the production network on the basis that we do serve sites around the world. > > Do you use ndmp? Still waiting for it to mature and appear in products we feel happy to use and rely upon. > > How many streams of backup do you use in parallel? One per filer, cos we've only *currently* got one device on each filer. > > What speed/rate do you backup (GB/Hr) per drive? About 2.3MB/sec. (I reduced the numbers to that dimension and can't recall what they were in GB/hr - I'm sure you can work them out). > > Do you utilize more than one tape drive for backups? One drive per box though we're looking at two or more per box in future. > > Have you needed to restore a file system? Sadly yes, way back a year or so ago on an older box on an older release of the software we suffered a catastrophic loss of a box. It took me a long time to restore the bits we needed because I was, at the time, reduced to using ufs_restore from a Sun over the network. Still managed 1MB/sec though. Otherwise I just smugly cp things out of snapshots. > > If so, how long did it take you (GB/Hr) ? Days. > > Were you able to retore in parallel streams? Nope. > > > Since the list of questions came out long, > I will be more than happy to call you > and poll you verbally on the above (send me your number). > I will even be more than happy to post the > statistical results of those willing to participate. I hope people chip in - it'll be interesting to see if backups on filers has become customer-target no 1, now that there're so many good things in DOT5 to play with.

1 0

Re: Restoring A LOT from snapshot
by sirbruce＠ix.netcom.com 22 Apr '98

22 Apr '98

On 04/22/98 10:00:56 you wrote: > Solution 4: boot from floppy, use prev_cp and wack? > Has anyone had any experience with 'prev_cp' and 'wack'? > >We're leaning at #4. Any comments would greatly be appreciated! If support says #4 will work, and they'll "support" it, then I would choose that method myself. Bruce

1 0

Re: backup times
by Mark D Simmons 22 Apr '98

22 Apr '98

> > On Mon, 20 Apr 1998, Stephen Manley wrote: > > > > > Depends. What version of ONTAP are you running right now that you see > > > this behavior? > > > > sorry, good question. 4.3.1D5, and possibly also with 4.3.3. > > > > The basic data I know, Data ONTAP 5.0 runs significantly faster (50%?) than 4.3. > Furthermore, it should not crank at 100% CPU for minutes/hours at a time. > We've gotten to the point here that people don't really notice an effect > when we run dumps, anymore. > > Also, if I could ask a survey question of people (if this is not an acceptable > forum for this, just flame me ;) How do people out there use dump? > > 1) rsh to local drive. Currently our only backup possibility, it's working fairly well for now but as I've mentioned previsouly we're looking to seriously load our filers (before getting some more 8) and the backup times are looking daft! Specially as we do full backups every night, and will continue to do so until such time as we have a sterling installation of something like Veritas such that incrementals become manageable and reliable. > 2) rsh to remote drive. Why? streaming across the main network is a bad idea (tm) and CIFS put paid to the utility of the technique anyhow. > 3) NDMP to local drive. This is what I want to move toward, using tape-drives attached to the filer (ie where the data lives) and control across the network to a Veritas/other control box, where the indexes get kept. > 4) NDMP to remote drive. Doh! Only when we've got our dedicated Gigabit Ethernet backup network in place, which current estimates place around the next ice-age. > 5) I don't. > > Inquiring engineering minds are curious to know. :) Forgive my sarcasm in the above - you wouldn't believe the angst the word 'backup' causes around here at times. Plus I think I got out the cynical side of my bed this morning. > > Sm

2 1

Jump to page:

2025

2024

2023

2022

2021

2020

2019

2018

2017

2016

2015

2014

2013

2012

2011

2010

2009

2008

2007

2006

2005

2004

2003

2002

2001

2000

1999

1998

1997

toasters