Thanks to everyone who has responded to my questions about the Polycenter
Save and Restore software. We're probably going to get it; we have a few
questions that our DEC rep is going to find answers for.
I've included all the responses I received to date on the software below.
Thanks again,
Mike
--------------------------------------------------------------------------
Date: Fri, 24 Mar 95 07:33:47 +0100
To: iglesias_at_draco.acs.uci.edu
From: phj06_at_bk.dk (Paul Hostrup-Jessen)
Subject: Re: Polycenter Save and Restore questions
Hello Mike,
> We're considering Polycenter Save and Restore to do our backups. We currentl
y
> back up about 110 systems around campus with homegrown scripts, but it
> is becoming too unwieldy to continue doing the saves that way. For those
> of you running the above software, I have some questions:
>
> 1. The documentation states that the database consumes about 10% of what
> is saved. If this is true, and we're saving at least 1gb per system
> on fullsaves, that's a lot of disk space. How are you handling that?
We use NetWorker ("Polycenter Save and Restore" is Digital's name for it), too,
and are quite happy about it. The 10% is a bit on the high side. It really
depends on which kind of directives/schedules you use for your backups. Each
file entry in the database takes up about 200 kByte. The idea is to limit the
amount of full backups to for instance once per week while doing incremental
backups on the remaining days. The size of the database also depends on your
choice of browsing/retrieving policy (how far back you want to be able to
restore your files). The database can grow considerably if you don't find the
right balance.
However, the partition you choose to install NetWorker on should be quite
large. By default the program installs on /usr, but you can choose a completely
different partition at the time of installation. The database can have a
maximum size of 1 GByte!
> 2. Have you ever had to restore the root and/or /usr partition of a system
> that has had a disk crash on those partitions? How hard was it?
Not yet! However, we do disk mirroring of the / and /usr partition and even
keep a spare system disk in the safe - just to be sure! We use StorageWorks
which makes it easy to swap a disk.
> 3. We're going to be running the client software on a variety of systems:
> Alphas running Digital Unix, Suns running SunOS & Solaris 2, SGIs running
> IRIX 4.x & 5.x, DECstations running Ultrix 4.x. Have you had any problem
s
> with the clients on those systems?
We haven't tried that yet, but we will soon find out. We also have Suns, HPs
and other DECs as well as a huge installation of Windows NT Advanced Server
clients.
Best regards
Paul Hostrup-Jessen
Department of Electronic Data Processing
Unix system administration
Bruel & Kjaer A/S
DK-2850 Naerum
DENMARK
email: phj06_at_bk.dk
tel. : +45 45 80 78 55 + 2433 (direct dial from touch-tone telephones)
fax : +45 42 80 14 05
Date: Fri, 24 Mar 95 08:12:51 +0100
To: iglesias_at_draco.acs.uci.edu
From: furrer_j_at_vptt.ch
Subject: RE: Polycenter Save and Restore questions
Hi Mike,
> 1. The documentation states that the database consumes about 10% of what
> is saved. If this is true, and we're saving at least 1gb per system
> on fullsaves, that's a lot of disk space. How are you handling that?
It depends on your retention and browse policy.
> 2. Have you ever had to restore the root and/or /usr partition of a system
> that has had a disk crash on those partitions? How hard was it?
To restore any disk is quite easy. To restore the / or /usr you have to install
first the client part of DECnsr on the client and then the recover is very
comfortable.
> 3. We're going to be running the client software on a variety of systems:
> Alphas running Digital Unix, Suns running SunOS & Solaris 2, SGIs running
> IRIX 4.x & 5.x, DECstations running Ultrix 4.x. Have you had any problem
s
> with the clients on those systems?
We have experiance with OSF1/2.x, ULTRIX 4.x and SunOS 4.1.2 and there is no
problem. Now we have orderd the client for Windows-NT.
> Any other things that you can think of would be appreciated.
>
+------------------+-------------------------------+---------------------------
+
| Jonas Furrer | Telefon : +41 31 338 94 40 | E-Mail : furrer_j_at_vptt.ch
|
| GD PTT RC 443 | Telefax : +41 31 338 44 10 | X-400 : c=ch; a=arcom
|
| Speichergasse 6 | Pager : +41 40 39 99 03 | p=swissptt; i=jfu
|
| 3030 Bern | | s=furrer; g=jonas
|
+------------------+-------------------------------+---------------------------
+
Date: Fri, 24 Mar 1995 09:16:27 +0200 (EET)
To: Mike Iglesias <iglesias_at_draco.acs.uci.edu>
From: Harald Lundberg <hl_at_tekla.fi>
Subject: Re: Polycenter Save and Restore questions
In-reply-to: <199503232143.AA04454_at_draco.acs.uci.edu>
On Thu, 23 Mar 1995, Mike Iglesias wrote:
>
> 1. The documentation states that the database consumes about 10% of what
> is saved. If this is true, and we're saving at least 1gb per system
> on fullsaves, that's a lot of disk space. How are you handling that?
It depends a lot on how *many* files you have, not how big they are. We
have a lot of big files, I backup 50Gb/week (full) + incrementals (once a
day), and my index is 1.5 Gb. I keep the online index for 1 month. If you
shorten this time, the index will shrink.
>
> 2. Have you ever had to restore the root and/or /usr partition of a system
> that has had a disk crash on those partitions? How hard was it?
No experience. Actually I don't even backup system disks. I just keep
copies of the (very few) files I've modified.
>
> 3. We're going to be running the client software on a variety of systems:
> Alphas running Digital Unix, Suns running SunOS & Solaris 2, SGIs running
> IRIX 4.x & 5.x, DECstations running Ultrix 4.x. Have you had any problem
s
> with the clients on those systems?
I have 2 servers, Alpha and Solaris 2.3. My clients are Ultrix, Alpha, HP,
Solaris. No problems.
>
> Any other things that you can think of would be appreciated.
If you can afford it, buy a *HUGE* jukebox. I have an HP1553A, six
slot DAT, and it barely fits my full week's backups. It's very nice,
my users can retrieve the last versions of their files without bothering
me. I'm going to buy a stand-alone dat for recoveries, then I don't have
to unload the jb for a single recover.
>
>
> Mike Iglesias Internet: iglesias_at_draco.acs.uci.edu
--harald
Harald Lundberg <hl_at_tekla.fi>;Tekla Oy,Koronakatu 1,FIN-02210,ESPOO,FINLAND
tel +358-{0-8879449work,0-8039489fax,0-8026752,11-728013res,50-5578303mob)
Date: Fri, 24 Mar 1995 09:48:52 +0100
To: Mike Iglesias <iglesias_at_draco.acs.uci.edu>
From: "Didier Wagenknecht, Security Officer, ADM-SIC, EPFL" <wagen_at_slalpha1.
***epfl.ch>
Subject: Re: Polycenter Save and Restore questions
In-reply-to: Your message of "Thu, 23 Mar 1995 13:43:55 MET."
<199503232143.AA04454_at_draco.acs.uci.edu>
Hi Mike,
to answer your questions:
1. I just did a du -k on the /nsr directory, it gave me 201363 k for
9 clients, each with =>1GB of files. In fact, the index grows not
because of the size of the disks, but with the number of files saved,
which is fairly different.
2. Touch some wood, it never happened here. But some day, I will crash
the disk and restore it, just in case...
3. The same for me here. Any platform except Macintoshes, no problem.
The only problem I had was with the DNS, and a little incomprehension
of the mecanism used to get the name of the client, or to return the
name of the server: the name sometimes is fully qualified, sometimes
not. I finally made some aliases in NSR, to have both names, and
no problems since that time.
Regards,
Didier
_____________________________________________________________
Didier Wagenknecht, Service Informatique Central (SIC/SL)
Ecole Polytechnique Federale de Lausanne
CH-1015 Lausanne (Switzerland)
E-mail : wagen_at_sic.adm.epfl.ch
Phone : (+41.21) 693.22.18, Fax : (+41.21) 693.22.20
Date: Fri, 24 Mar 1995 10:58:27 +0100
To: Mike Iglesias <iglesias_at_draco.acs.uci.edu>
From: Hellebx Knut <bgk1142_at_bggfu2.nho.hydro.com>
Subject: Re: Polycenter Save and Restore questions
In-reply-to: Mike Iglesias <iglesias_at_draco.acs.uci.edu>
"Polycenter Save and Restore questions" (Mar 23, 1:43pm)
Regards,
We use Legato NetWorker (same product, modified slightly by DEC). We have
installed Alpha machines as clients to our NetWorker server and in addition
we have IRIX, Solaris, SunOS, HP/UX, ULTRIX, AIX clients. All works OK.
The size of the indexdatabases is not by far as big as 10% the size of
what's saved. The most spaceconsuming client we have is a RS/6000 machine
running ECLIPSE that generates a bunch of files (the indexdatabase grows as
a function of #files, not so much the size of them) and that client
consumes about 3% of the space saved (full backup: ~7GB, indexdatabase: 217
MB)
We have not a / or /usr recovery on any machines yet (thank God) but I
think you will be well off following the manuals (at least the Legato
NetWorker manuals are clear at that point ;-) )
Also note that in near future archiver facilities will be available for
DECnsr.
Finally, an advice: Get FAST tapedevices and a changer (DLT or something
like that) and a LOT of memory (at least ~128MB) on the server if you are
going to serve as much as 110 clients. And do see to that your network is
not suffering from any deseases. This could cause headaches when you're
trying to figure out why the backups went wrong.
--
******************************************************************
* Knut Helleboe | DAMN GOOD COFFEE !! *
* Norsk Hydro a.s | (and hot too) *
* Phone: +47 55 996870, Fax: +47 55 996342 | *
* Pager: +47 96 500718 | *
* E-mail: Knut.Hellebo_at_nho.hydro.com | Dale Cooper, FBI *
******************************************************************
Date: Fri, 24 Mar 95 13:18:41 +0000
To: iglesias_at_draco.acs.uci.edu
Cc: j.studnicka_at_ic.ac.uk
From: j.studnicka_at_ic.ac.uk
Subject: DECnsr
Mike,
here at Imperial we backup about 80 clients now (departmental
servers) and they are the same systems as at your place ie (Digital, SGI
and Sun). We have been running DECnsr service for about two years using
two DECstations 5000/133 as the servers. While the service has became
increasingly popular in the College we have now reached the stage when
the two servers can not cope and we are considering to upgrade the
servers to Alphas plus also to introdude third server for research groups.
At present 80 clients are organised into 10 groups, 5 groups per server.
Total client disc space is 247Gbytes holding 146Gbytes of data.
1.the figure of 10% for total backup data is roughly correct plus you
will need about 3% of scratch space to manipulate the database.
We have ~7.5Gbytes of discs on each server and they are full!
2.I'll tempt the fate and say we did not have to do it. However what we did
as a test was to restore / and /usr of a client. The easiest thing to do
this is to mount a disc onto any running client and do restore to the
mounted disc.
3.This we found to be a problem. DECnsr can (and by default does) compress
clients files before backing them up. Compression is very resource
demanding and on "old" machines (DEC 5000/200, SUN Sparcs, yes we have
some of these as well) their CPU simply can not cope. (CPU idle time
permanenty set to 0, hence your backup will take hours. And if you run
parallelism you can end up with more then one compression processes on
the machine.
We have also found that we need about 80Mbytes of memory on the servers
to prevet excessive swapping during the backup sessions.
Good luck and regards
Jiri
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>
Jiri V. Studnicka
Centre for Computing Services, phone 071-594 6942
Imperial College of Science Technology & Medicine, fax 071-594 6958
London SW7 2BX, England email j.studnicka_at_ic.ac.uk
>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
>
Date: Fri, 24 Mar 95 08:22:28 -0600
To: Mike Iglesias <iglesias_at_draco.acs.uci.edu>
From: jldetter_at_cca.rockwell.com
Subject: Re: Polycenter Save and Restore questions
>We're considering Polycenter Save and Restore to do our backups. We currently
>back up about 110 systems around campus with homegrown scripts, but it
>is becoming too unwieldy to continue doing the saves that way. For those
>of you running the above software, I have some questions:
>
>1. The documentation states that the database consumes about 10% of what
> is saved. If this is true, and we're saving at least 1gb per system
> on fullsaves, that's a lot of disk space. How are you handling that?
>
>2. Have you ever had to restore the root and/or /usr partition of a system
> that has had a disk crash on those partitions? How hard was it?
>
>3. We're going to be running the client software on a variety of systems:
> Alphas running Digital Unix, Suns running SunOS & Solaris 2, SGIs running
> IRIX 4.x & 5.x, DECstations running Ultrix 4.x. Have you had any problems
> with the clients on those systems?
>
>Any other things that you can think of would be appreciated.
>
>
>Mike Iglesias Internet: iglesias_at_draco.acs.uci.edu
>University of California, Irvine phone: (714) 824-6926
>Office of Academic Computing FAX: (714) 824-2069
>Central Computing Support
Mike
We are using Legatos Networker to backup roughyl 350 workstations.
We have 2-200 license servers that we run on two sparcs. We are
primarily a DEC house but bought networker from Legato because it
was not available from DEC when we needed it.
It does take disk space, however this is up to you. The amount of
disk space is proportional to the numbers of files that you backup
and the length of time that you want the data to be stored on-line.
We save our data for 6 months and require around 10 Gb per server
to keep this data. We onlt backup selected directories on each
machine so I can answer your second question.
As far as clients we currently run:
ULTRIX
OSF1
SunOS
Solaris
IRIX
HP-UX
AIX
Unixware
If you have any additional questions I would be glad to help. You
might also try Alan Hanson (abhanson), he is the one that actually
does most of our backups here at Rockwell.
Jeff
--------
______________________________________________________________________
| Jeff Detterman (319) 395-2274 Rockwell International |
| Unix Network Administrator Collins Commercial Avionics Div |
| jeff_at_cca.rockwell.com Cedar Rapids, Iowa 52498 |
----------------------------------------------------------------------
Date: Fri, 24 Mar 95 08:59:40 -0800
To: Mike Iglesias <iglesias_at_draco.acs.uci.edu>
From: Jon Forrest <forrest_at_postgres.Berkeley.EDU>
Subject: Re: Polycenter Save and Restore questions
In-reply-to: Your message of "Thu, 23 Mar 95 13:43:55 PST."
<199503232143.AA04454_at_draco.acs.uci.edu>
> 2. Have you ever had to restore the root and/or /usr partition of a system
> that has had a disk crash on those partitions? How hard was it?
I'm not running it but people I know and respect are. They tell me
they get around this problem by using regular dump to do a level 0
dump of / and /usr every so often. This makes it easier in case
of disaster.
Jon
Received on Fri Mar 24 1995 - 15:51:28 NZST