01-06-2010 07:37 AM
I have this question. In our tape fabric envrionment, where inherently the infrastructure is 1 initiator to multiple targets, and in this kind of environment is there still a need to have "one initiator to one target" practiced?
How are you guys dealing with this, how do you zone your master server to tape devices?
01-07-2010 09:24 PM
I note you have post a similary question here in the past.
is this the same issue ?
what kind of things do you intend to realize ?
01-07-2010 09:37 PM
Is my thinking correct on the linked post?
This question is concerning the same environment, but not the same question. My current question concerns more with zoning best practice, specifically how many initiators, targets should exist in a zone? 1:1, 1:n, n:n?
If we want to proceed with a one initiator to multiple targest zoning, ie. 1:n, how should we name our zones? For example, if one netbackup master server connects to a tape silo with multiple panels and each panel has multiple tape drives, how should this be represented in the zone name? or it shouldn't be?
01-08-2010 07:10 AM
In your original (linked) post, your question seems to be "does FSPF work?". The answer is yes!
For the current post, I would say use 1:n zoning with a limit of 4 or 5 targets per zone to contain RSCN propagation. The zone name would indicate the initiator. If the NBU master needs access to more than 5 target drives, then split them up into several zones (by panel, e.g.).
On the other hand, there is nothing stopping you from using 1:1 zoning if you're really paranoid.
Just my opinion,
01-08-2010 07:54 AM
Thanks for the reply. I have all the intention to go with 1:n model considering our fabric inherent properties.
However, Alastir, I still think you are too restrictive on your 1:4-5 consideration. Are there any specific concerns for RSCN "broadcast" you are worried about?
01-11-2010 12:13 AM
If a drive goes down for whatever reason, every other drive in the zone will have to PLOGI and the more entries in the zone, the more activity on the SAN and the longer it will take to quiet down again. Are you convinced your drives will support gracefully this disturbance when in full backup?
I would minimise the risk and think 4 or 5 targets is sufficient. Of course, if I had a robot with 12 drives (6 on each fabric) I would extend that limit to 6. It's a question of being reasonable.