SUMMARY: nsr and indexes

From: Jason Neil <jan_at_nzxray.galen.co.nz>
Date: Tue, 13 Feb 1996 14:30:45 -0700 (PDT)

My appretiation to all those who took the time to reply,

lucien HERCAUD
Hellebo knut
Harald Lundberg
Paul E. Rockwell

Thankyou.

There was question as to why I wanted a client initiated save, the
answer was because I was shutting down a data base and creating clones
etc... which all have to be controlled by the client in question.

I will tail the reply's on the back, but here's the guts of the replies.

1. Set up backup group and use the savegroup command as
oppossed to the save command.

2. Retention policies are controlled via the client definition on the
NSR server. You set that with the browse and retention policies for the
client.

3. Incr backups, and short retention policies keep online indexes down
to a minimum.


Replies follow -----

>From hl_at_tekla.fiTue Feb 13 14:29:16 1996
Date: Thu, 8 Feb 1996 08:56:32 +0200 (EET)
From: Harald Lundberg <hl_at_tekla.fi>
To: Jason Neil <jan_at_nzxray.galen.co.nz>
Subject: Re: nsr and indexes

On Thu, 8 Feb 1996, Jason Neil wrote:

> Hi
>
> We have a alpha running DU3.2c with NSR311, and are having a
> great deal of trouble keeping our online indexes to a managable size.
> The backups are instigated from a shell script using the save command.
> Retention policies have been defined using the GUI software, but I have
> the feeling that the unix commands bypass that somehow.
Is there a specific reason why you don't use a normal automagic savegroup?
>
> My questions are:
>
> Is save the best command to use for backing up filesystems using nsr?
> (note: vdumps of the whole system are done monthly)
I use automatic savegroups with a few special directives (like skipping
.o and .a files). I think this is the way it was meant to be...
>
> Do you need to specify some option on the unix command to tell networker
> to use a particular retention policy.
No. The retention police is set up for each client in the client setup.
>
> Basically, how would I keep the on-line on-line index's down to say a
> week or month?
You can specify a policy of your own: 11 days or whatever you wish. This
policy can be used as a retension police (this is how long the tape will
be considered 'valuable'), OR, as a browse policy. The browse policy
says how long stuff will be kept in the online index. Default is one
month.
>
> Any other tips on nsr for DU would be appreciated as well.
>
It seems to me that you haven't been using the graphical interface,
'networker'. I'd suggest you have a real thorough look at it. The
'nsradmin' way is not as intuitive.

> Thanks in advance,
>
--harald
Harald Lundberg <hl_at_tekla.fi>;Tekla Oy,Koronakatu 1,FIN-02210,ESPOO,FINLAND
tel +358-{0-8879449work,0-8039489fax,0-8026752,11-2418013res,50-5578303mob)

>From Knut.Hellebo_at_nho.hydro.comTue Feb 13 14:29:24 1996
Date: Thu, 8 Feb 1996 08:28:37 +0100
From: Hellebo Knut <Knut.Hellebo_at_nho.hydro.com>
To: Jason Neil <jan_at_nzxray.galen.co.nz>
Subject: Re: nsr and indexes

On Feb 8, 12:38pm, Jason Neil wrote:
> Subject: nsr and indexes
> Hi
>
> We have a alpha running DU3.2c with NSR311, and are having a
> great deal of trouble keeping our online indexes to a managable size.
> The backups are instigated from a shell script using the save command.
> Retention policies have been defined using the GUI software, but I have
> the feeling that the unix commands bypass that somehow.

What do you mean by "I have the feeling that the unix commands bypass that
somehow" ? Why don't you let NetWorker run your backups run automatically by
setting up a reasonable schedule from the GUI ?

>
> My questions are:
>
> Is save the best command to use for backing up filesystems using nsr?
> (note: vdumps of the whole system are done monthly)

No. Set up backupgroups and schedules and start them either automatically
from NetWorker or by doing 'savegroup'

>
> Do you need to specify some option on the unix command to tell networker
> to use a particular retention policy.

No. Just see to that you have a reasonable retention policy/browsetime. The
browsetime determines the indexsize. A larger browsetime means bigger
indexfiles.

>
> Basically, how would I keep the on-line on-line index's down to say a
> week or month?
>
> Any other tips on nsr for DU would be appreciated as well.
>
> Thanks in advance,

You're welcome

>
>-- End of excerpt from Jason Neil



-- 
      ******************************************************************
      *         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    *
      ******************************************************************
>From Lucien HERCAUD_at_paribas.comTue Feb 13 14:29:38 1996
Date: Thu, 8 Feb 96 14:17:00 PST
From: "Lucien HERCAUD"_at_paribas.com
To: jan_at_nzxray.galen.co.nz
Subject: Re: nsr and indexes
     From my own experience:
     
     If you need client initiated scheduling of the backup operation,
     better say on the client :
        "rsh nsrhost savegroup my_client_group [-c my_client] [-l incr]"
     than doing a "save" on the client itself.
     
     The "save" solution has 2 disadvantages:
     
     1. does no imply the saveindex for the client's index on the server -which 
     savegroup does-.
     
     2. You need to specify the [-T "time"] option as taken from the last time 
     "save" was done - see the man page. Otherwise you save all the files as if 
     a full backup is asked, even if "-l incr" is specified.
     
     Try the savegroup, maybe the "incr" will save you some space in the 
     indexes.
     
     If not, you need to play with the BROWSE policy. The shorter the BROWSE is, 
     the smaller the online indexes will get. Also, you can manually go in and 
     PURGE the older versions for each filesystem of each client (see the 
     "indexes" menu).
     
     
                                Lucien HERCAUD
>From rockwell_at_rch.dec.comTue Feb 13 14:29:49 1996
Date: Mon, 12 Feb 96 11:56:53 -0500
From: "Paul E. Rockwell" <rockwell_at_rch.dec.com>
To: Jason Neil <jan_at_nzxray.galen.co.nz>
Subject: Re: nsr and indexes 
On Thu, 8 Feb 1996 12:38:15 -0700 (PDT)  jan_at_nzxray.galen.co.nz wrote:
> Hi
> 
> We have a alpha running DU3.2c with NSR311,  and are having a 
> great deal of trouble keeping our online indexes to a managable size.  
> The backups are instigated from a shell script using the save command. 
> Retention policies have been defined using the GUI software,  but I have 
> the feeling that the unix commands bypass that somehow.
Is there a special reason that you're using NSR's save as opposed to
letting NSR schedule the backups? For example, are you running
a database that you're shutting down with the script, then making a fileset
clone with AdvFS?
> Is save the best command to use for backing up filesystems using nsr?  
> (note: vdumps of the whole system are done monthly)
Depends on what you're doing. You may be doing redundant vdumps if you're
doing the whole system monthly. You should only need to vdump the root,
usr, and wherever the NSR indexes are residing. The rest can be handled
quite nicely with NSR.
> Do you need to specify some option on the unix command to tell networker 
> to use a particular retention policy.   
Retention policies are controlled via the client definition on the NSR server.
You set that with the browse and retention policies for the client.
> 
> Basically,  how would I keep the on-line on-line index's down to say a 
> week or month?
What's your backup schedule look like? Are you doing a full backup each time?
That's a quick way to bloat the size of the NSR indicies.
To keep the size of the on-line indexes down:
1) Choose an appropriate browse time. Once all save sets in a backup cycle (the backup 
cycle is a full, followed by all incrementals/level n backups until the next full) 
have passed the browse time as defined to your client, then the entries will be
removed from the client index (but the save sets will still be in the  media
index, and recoverable via scanner). So, if your backup cycle is one week, and
your browse policy is 1 week, then you can have up to 2 weeks of data in the on-line
indexes. See the "addendum.ps" file that's released with NSR (this is the Networker
Addendum manual" for a more in-depth discussion of the interrelationships among
browse, retention, and backup cycles.
2) Relating to 1), create an appropriate backup schedule.
3) Relating to 2): Use incremental backups liberally.
--------
+---------------------------+tm		Paul E. Rockwell
|   |   |   |   |   |   |   |		Northeast Region SBU Technical Support
| d | i | g | i | t | a | l |		Digital Equipment Corporation
|   |   |   |   |   |   |   |   	500 Enterprise Drive
+---------------------------+		Rocky Hill, CT 06067
Internet: rockwell_at_rch.dec.com		Phone: (860)258-5022
 
  -------------------------------------------------------------------
< Jason Neil                            Contact Details               >
< Technical Co-ordinator                Phone: 0064 9 3771780         >
< Galen Group Ltd                       Fax:   0064 9 3771787         >
< Auckland, New Zealand                 email: jan_at_nzxray.galen.co.nz >
  -------------------------------------------------------------------
Received on Tue Feb 13 1996 - 03:20:55 NZDT

This archive was generated by hypermail 2.4.0 : Wed Nov 08 2023 - 11:53:46 NZDT