Zones and RSCN
 

Hello,

I'm new to SANs so please excuse my ignorance. I've been reading
about zoning and RSCNs. From what I've read RSCNs are sent to all
members of a zone after a change in the port status occurs on a port
in that zone. For example, if all servers are in the same zone, and
one of them is rebooted, an RSCN is sent to all the other hosts in
that same zone to notify them of the event. At this time are the
other hosts in the zone "knocked offline" termporarily while they do
name server refresh? How can you prevent this? I've read a post that
somebody said to create one zone per port, this way it won't affect
any other hosts. What can you do if you have a backup type
environment where all hosts need to speak to one host (the backup
server)?

-TIA

I'm new to SANs so please excuse my ignorance. I've been reading
about zoning and RSCNs. From what I've read RSCNs are sent to all
members of a zone after a change in the port status occurs on a port
in that zone. For example, if all servers are in the same zone, and
one of them is rebooted, an RSCN is sent to all the other hosts in
that same zone to notify them of the event. At this time are the
other hosts in the zone "knocked offline" termporarily while they do
name server refresh? How can you prevent this? I've read a post that
somebody said to create one zone per port, this way it won't affect
any other hosts.

Yes, thats true. You should define a zone allways with only one FC-HBA
member.

What can you do if you have a backup type
environment where all hosts need to speak to one host (the backup
server)?

Sorry, I don't understand this question. Why should the client hosts speak
to the backup server over FC? Most of this solutions I know talk via IP for
Information/Metadata Interchange. What specific solution is that you speak
about?

René

-TIA

I beleive the OP is referring to backing up to a media server over the
fiber channel to (possibly) fiber channel drives.

In this case you would create a seperate zone that includes all backup
clients and the server. When it's time to do backups move all hosts
into this zone. This requires some scripting on the switch side but I
know of at least one person who does this exact config.

Obviously the backup zone needs to include any members that the client
communicated with in the non-backup zone.

~F

On Mon, 20 Oct 2003 16:47:08 +0200, "Rene Köhnen-Wiesemes"
wrote:

I'm new to SANs so please excuse my ignorance. I've been reading
about zoning and RSCNs. From what I've read RSCNs are sent to all
members of a zone after a change in the port status occurs on a port
in that zone. For example, if all servers are in the same zone, and
one of them is rebooted, an RSCN is sent to all the other hosts in
that same zone to notify them of the event. At this time are the
other hosts in the zone "knocked offline" termporarily while they do
name server refresh? How can you prevent this? I've read a post that
somebody said to create one zone per port, this way it won't affect
any other hosts.

Yes, thats true. You should define a zone allways with only one FC-HBA
member.

What can you do if you have a backup type
environment where all hosts need to speak to one host (the backup
server)?

Sorry, I don't understand this question. Why should the client hosts speak
to the backup server over FC? Most of this solutions I know talk via IP for
Information/Metadata Interchange. What specific solution is that you speak
about?

René

-TIA

Sounds interesting. But personally I've never seen such kind
of configuration. What I have implemented up to now looks
different:
Veritas Netbackup uses a central Backup Server and the Media Server
option for these servers which holds huge amount of data. These Media
Servers have direct connection to the tapes in a tape library like the
Backup Server. The trick is that tapes are shared between Backup Server
and Media Server. All other servers can be backup via LAN.
Nevertheless, you can configure zones that contains only one hba per zone.

Legato Networker works the same, only nomenclatura is different, the Media
Servers are called Storage Nodes, and the licence you should have is DDS
(Dynamic Device Sharing). As obove, each fc hba should have his private
zone.

René

SAN configuration rule of
"Faeandar" schrieb im Newsbeitrag
...
I beleive the OP is referring to backing up to a media server over the
fiber channel to (possibly) fiber channel drives.

In this case you would create a seperate zone that includes all backup
clients and the server. When it's time to do backups move all hosts
into this zone. This requires some scripting on the switch side but I
know of at least one person who does this exact config.

Obviously the backup zone needs to include any members that the client
communicated with in the non-backup zone.

~F

On Mon, 20 Oct 2003 16:47:08 +0200, "Rene Köhnen-Wiesemes"
wrote:

I'm new to SANs so please excuse my ignorance. I've been reading
about zoning and RSCNs. From what I've read RSCNs are sent to all
members of a zone after a change in the port status occurs on a port
in that zone. For example, if all servers are in the same zone, and
one of them is rebooted, an RSCN is sent to all the other hosts in
that same zone to notify them of the event. At this time are the
other hosts in the zone "knocked offline" termporarily while they do
name server refresh? How can you prevent this? I've read a post that
somebody said to create one zone per port, this way it won't affect
any other hosts.

Yes, thats true. You should define a zone allways with only one FC-HBA
member.

What can you do if you have a backup type
environment where all hosts need to speak to one host (the backup
server)?

Sorry, I don't understand this question. Why should the client hosts
speak
to the backup server over FC? Most of this solutions I know talk via IP
for
Information/Metadata Interchange. What specific solution is that you
speak
about?

René

-TIA

Some people want to get backup traffic off the LAN if possible so they
use either SAN media servers or SSO over fiber (Veritas specific lingo
here). With this you can move all your backup traffic over the fiber
and keep the LAN available for user access.

You can only have one hba in one zone *at a time*. There is nothing
to prevent one hba being in multiple inactive zones concurrently.

Like I said, it takes some scripting at the switch but you can easily
move the hba into a "backup" zone.

~F

On Mon, 20 Oct 2003 21:57:39 +0200, "Rene Köhnen-Wiesemes"
wrote:

Sounds interesting. But personally I've never seen such kind
of configuration. What I have implemented up to now looks
different:
Veritas Netbackup uses a central Backup Server and the Media Server
option for these servers which holds huge amount of data. These Media
Servers have direct connection to the tapes in a tape library like the
Backup Server. The trick is that tapes are shared between Backup Server
and Media Server. All other servers can be backup via LAN.
Nevertheless, you can configure zones that contains only one hba per zone.

Legato Networker works the same, only nomenclatura is different, the Media
Servers are called Storage Nodes, and the licence you should have is DDS
(Dynamic Device Sharing). As obove, each fc hba should have his private
zone.

René

SAN configuration rule of
"Faeandar" schrieb im Newsbeitrag
.. .
I beleive the OP is referring to backing up to a media server over the
fiber channel to (possibly) fiber channel drives.

In this case you would create a seperate zone that includes all backup
clients and the server. When it's time to do backups move all hosts
into this zone. This requires some scripting on the switch side but I
know of at least one person who does this exact config.

Obviously the backup zone needs to include any members that the client
communicated with in the non-backup zone.

~F

On Mon, 20 Oct 2003 16:47:08 +0200, "Rene Köhnen-Wiesemes"
wrote:

I'm new to SANs so please excuse my ignorance. I've been reading
about zoning and RSCNs. From what I've read RSCNs are sent to all
members of a zone after a change in the port status occurs on a port
in that zone. For example, if all servers are in the same zone, and
one of them is rebooted, an RSCN is sent to all the other hosts in
that same zone to notify them of the event. At this time are the
other hosts in the zone "knocked offline" termporarily while they do
name server refresh? How can you prevent this? I've read a post that
somebody said to create one zone per port, this way it won't affect
any other hosts.

Yes, thats true. You should define a zone allways with only one FC-HBA
member.

What can you do if you have a backup type
environment where all hosts need to speak to one host (the backup
server)?

Sorry, I don't understand this question. Why should the client hosts
speak
to the backup server over FC? Most of this solutions I know talk via IP
for
Information/Metadata Interchange. What specific solution is that you
speak
about?

René

-TIA

Faeandar wrote in message . ..
Some people want to get backup traffic off the LAN if possible so they
use either SAN media servers or SSO over fiber (Veritas specific lingo
here). With this you can move all your backup traffic over the fiber
and keep the LAN available for user access.

You can only have one hba in one zone *at a time*. There is nothing
to prevent one hba being in multiple inactive zones concurrently.

Like I said, it takes some scripting at the switch but you can easily
move the hba into a "backup" zone.

~F

Faeandar,

I've seen many configurations where a single HBA is in multiple active
zones. For instance, you can have a single HBA zoned to two storage
targets, and create two zones for this purpose:

Zone 1: HBA_A - Storage_A
Zone 2: HBA_A - Storage_B

And beyond this, you could also have HBA_A zoned to, say, a SCSI-Fibre
Bridge for a tape library, if this host was a media server or was
using the Netbackup SSO or something to that extent:

Zone 3: HBA_A - Bridge_A

You could have all three of these zones active at the same time. You
would just have to ensure that you defined the persistent binding
appropriately on the host side. In other words, you might have
Storage_A on target 16, Storage_B on target 17, and Bridge_A on target
18.

I don't see any reason why anyone would want to write scripts to
change their zoning configurations on a daily basis -- that sounds
kind of risky to me. Especially considering the fact that if you were
_MOVE_ the HBA out of its "production" zone and into a "backup" zone
on a nightly basis, you would lose access to your production volumes..

- Sean

Hmm, good to know. I wonder if I just got bad info or if this was
case in earlier days. I have never done this personally so I can't
say for certain. The person I know who did do it would move the HBA
into a backup zone at night rather than put all the clients in the
same zone with each other and the backup server. Security perhaps? I
know he managed daytime access somehow as well, just don't remember
the specifics.

In any case OP, it can be done without going over the LAN.

~F

On 21 Oct 2003 09:19:31 -0700, (Sean Cummins) wrote:

Faeandar wrote in message . ..
Some people want to get backup traffic off the LAN if possible so they
use either SAN media servers or SSO over fiber (Veritas specific lingo
here). With this you can move all your backup traffic over the fiber
and keep the LAN available for user access.

You can only have one hba in one zone *at a time*. There is nothing
to prevent one hba being in multiple inactive zones concurrently.

Like I said, it takes some scripting at the switch but you can easily
move the hba into a "backup" zone.

~F

Faeandar,

I've seen many configurations where a single HBA is in multiple active
zones. For instance, you can have a single HBA zoned to two storage
targets, and create two zones for this purpose:

Zone 1: HBA_A - Storage_A
Zone 2: HBA_A - Storage_B

And beyond this, you could also have HBA_A zoned to, say, a SCSI-Fibre
Bridge for a tape library, if this host was a media server or was
using the Netbackup SSO or something to that extent:

Zone 3: HBA_A - Bridge_A

You could have all three of these zones active at the same time. You
would just have to ensure that you defined the persistent binding
appropriately on the host side. In other words, you might have
Storage_A on target 16, Storage_B on target 17, and Bridge_A on target
18.

I don't see any reason why anyone would want to write scripts to
change their zoning configurations on a daily basis -- that sounds
kind of risky to me. Especially considering the fact that if you were
_MOVE_ the HBA out of its "production" zone and into a "backup" zone
on a nightly basis, you would lose access to your production volumes..

- Sean

Just so everyone knows there are typically two behaviors of RSCNs and
zoning changes out there.

I remember working on a McData switch that had it zoning behavior
configureable. In one state it would send an RSCN to ALL zones when ANY
zoning changes are made. Then there was another state when it would only
send an RSCN to zones that actually changed. So if a change was made to
zone A, ONLY zone A gets an RSCN.

I think that both methods are right according to the spec. But I think
the second one is cleaner...

Lyle