Chris.Romano AT Lazard DOT com wrote:
>I am trying to decide upon 2 different setups.
>
>1) 2 separate Netbackup 5.1 servers, each with it's own library
>
>2) 1 master/media server + 1 media-only server (e.g. the master server has it's
>own silo and also controls a remote media server with it's own silo.
>
>Which configuration would be best for DR (if any) ?
>
>e.g....In option 1, I would have a surviving master and need to merge in the
>broken master's catalogs, clients, etc. into the surviving master.
>whereas in option 2, if I lost the master, I would need to make the media
>server
>into my new master, and import all catalog info, etc.over to it and make it
>into
>a master/media instead of just a media server.
>
>Which of the above 2 would be more complicated to do in a disaster?
>
>Thanks in advance for the advice.
>
>
>Chris.
>
>
>
>
>
>_______________________________________________
>Veritas-bu maillist - Veritas-bu AT mailman.eng.auburn DOT edu
>http://mailman.eng.auburn.edu/mailman/listinfo/veritas-bu
>
>
>
Are the 2 servers going to be on different sites? Im sure there are pros
and cons for each. If the master server is not specced to handle the
amount of data you are backing up or the amount of clients then you may
be better having a separate media server.
I however would go with option 1, and is currently how i run things
across 2 sites for a customer one of which is a DR site.
This is how.
Site1.
Master/Media server backs up around 20-25 clients over different
policies. I have a DR volume pool and In line tape copy or duplicate
backups on a daily basis which are sent off site. The normal backups are
retained on site in event of a restore being required. Each day the
catalouges from this master server are backed up to disk ( bpbackupdb
-dpath ) and ftpd as the image files to the DR site.
Site 2
Master /Media server backs up around 20 clients on the DR site ( which
is also a development site ). There is no offsite policy here as its a
DR site. I still backup the catalogues to disk as well as tape to
recover from the next step.
In Event of DR.
The critical thing is i have the master servers named the same across
the 2 sites, even though the clients are different names.
In event of DR i request the DR tapes from offsite ( written at site 1 )
and i wipe the catalogues from site 2 and bprecover the ones from site
1. After changing the robot device path ( because it was different ) and
stopping starting deamons, my site 2 now thinks its site 1. It also has
all policy, client information from site 1. This is obviously bad if
you wanted to restore from or backup a client from site 2, but in event
of a DR that is not my requirement.
As the hosts file still contains all hostnames for the site2 clients i
can restore to them. Therefore i start a restore on site2 which is now
the same as site 1, from a site 1 Webserver and i restore it to site 2
webserver. Im only restoring key directories not the whole machine. We
basically make the DR site function with the content of databases, apps
and web of site 1.
This has now worked in practice through 2 tests. After success i restore
the site 2 catalogues ( from my disk backup ) and site 2 is back to how
it was.
There may be better ways to do all this, but for the cost this to me is
good and works in practice. It may give some ideas for you to explore so
i hope it helps.
Cheers
Dave
|