Hello Aure,
You don't give many details, but I am doing a bit of research in that area.
(haven't tried it yet!) I assume you have a fiber attached library on your
networker server :)
Sounds like a "SAN" or "dedicated" storage node, if you are going to use DDS
and share the drives with the server.
You don't need to install a new jukebox on the cluster, but you do need the
storage node binaries as you say. To define DDS, you need to remove the
library from the networker server and re-configure it. It might be a long
time since you did it, or just upgraded it from an earlier version, but at
least from version 6.1x you are asked for each drive 1) if it is going to be
shared, 2) who is owning it. This means you may choose to dedicate a drive
to your cluster and save a DDS license, but then you don't use it fully and
I believe you need a shared jukebox license (please check it).
If you share them all, which is probably the most practical solution, you
just specify a second device path for the drive (you are prompted when you
declare it as shared) with the remote device form, i.e.
rd=sqlserv:\\.\Tape0. You can share the drive with as many storage nodes as
you like. Here you need to define the physical nodes (both), not the
virtual.
Your client definitions just need a few modifications, where your physical
nodes will use themselves as storage nodes -and then the nsrserverhost- and
your virtual nodes will point at phys_node1 and phys_node2. It seems that if
you change the virtual node to another physical node, you need to modify
this entry manually (or wait until it times out and attempts the next one.
There are some things to be careful about, like the SCSI reservation issue
which is a problem with DDS. Legato has posted a patch for 6.1.3 and I think
it is not ready for 7.0 yet, which adds a little option in your jukebox
configuration. The patch number is lgtpa48782, so you can read more about
it. The problem seems to be that other machines with access to the device
might send a SCSI reset on the bus and your backup is dropped, then someone
else requesting the drive might overwrite your tape - hence loss of backup
data. I hope someone else will give a better explanation :|
Personally, I think my option will be to dedicate a drive to such a cluster
pair and use a special pool for this backup until I hear this problem is
totally cleared. I also need to implement it soon and don't need the extra
risk. But it does depend on how many drives you have available.
Legato support goes still very careful about this kind of solution. It will
be very interesting to see what everyone here have to say.
Best regards,
Carolina
-----Original Message-----
From: Aurea Cruz [mailto:aurecruz AT YAHOO DOT COM]
Sent: 27. juni 2003 15:53
To: NETWORKER AT LISTMAIL.TEMPLE DOT EDU
Subject: clustered storage nodes installation
oted-printable by mailgw.proact.dk id PAA18855
Dear helpful forum,
I am turning my MSCS/SQL cluster (act/pass) into a SAN
storage node because it has gotten too large for my
busy net to bear a full backup of all servers, and the
SQL cluster is already talking to disks via fiber.
Assuming I deal with licensing the right way for
version 6.1.3 on solaris server, I am then doubting
about the procedure. Please tell me if this is
correct:
- where I'm really lost is on how do I declare this storage node, I will use
DDS, so should I
- I must reinstall the software on each node, reinstalling with storage node
binaries
- Must I register a cluster service? I did once for a networker server but
my clients didn't need to register anything. If so, then I have to move the
active node, right?
- the storagenodes value on the resource. Should it point at the virtual or
the physical clients? -should I remake the client definitions?
I have not seen many mails in the list talking about
this combination. Is anyone using it?
Thanks for any help !
Aure
--
Note: To sign off this list, send a "signoff networker" command via email
to listserv AT listmail.temple DOT edu or visit the list's Web site at
http://listmail.temple.edu/archives/networker.html where you can
also view and post messages to the list.
=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=*=
|
