Copy Lun to Lun

hmm ..i thought it was considered a bad practice to zone one initiator to two targets in one zone ?

EMC advises single initator (multiple target) zoning but they also acknowledge that mirrorview, sancopy setups are exception to this rule...

for hosts, SPs are the target but when mirrorview is setup, high order sp ports behave as target as well as intiator. they connect to other sp port as a scsi initiator and thus the target sp is able to register it.

nah, i would disagree,
single initiator zoning is a legacy of SCSI AFIK, in the old scsi world, there used to be only one initiator all the time on the scsi bus (scsi controller) and rest all behaved as targets (disks, CD-ROMs, Tapes). since that legacy is carried further in FC-SANs, single initiator zoning stays.

with multiple targets mean more than one device connected to the scsi cable, daisy chained as usual.

there are shared scsi buses available where multiple targets share the same bus and arbitrate between themselves. however the software (usually the OS driver) needed to be aware of the setup.

if you make one zone per initiator and target pair, you can run out of maximum zones limit in a zoneset very fast in a data centre.

You can disagree if you like, but these guys have been right every time we've had to use them, so I'm inclined to believe them when they say "Don't do it".

And we are using PowerPath. I am being told that PowerPath will not neccessarily distinguish multiple paths within a single zone when it comes to managing failover. It isn't just one engineer telling me this, so I'm not taking chances (and we don't have a storage test lab to play with this enough to make it worth trying in production).

Don't worry dynamox, we still do all our zoning (at least for CLARiiON connectivity) as single initiator. Every host gets four zones (or two for single HBA test hosts)), each zone has one HBA and one SP port.

This came straight from a couple of senior CLARiiON support techs. Failover and trespass can have problems when you configure it differently. In some circumstances putting multiple SP ports in a single zone can actually force a LUN trespass when a failover should have been possible. Not the end of the world, but why put more stress on the system than you have to.

Symms are a whole different ballgame though. You can zone multiple FA ports to a single HBA without affecting failover because a Symm uses an active/active configuration instead of the CLARiiON's active/passive config.

for me HBA is a initiator and SP port is (generally) a target. this is consistent to what i extracted from SCSI basics. anyone disagrees?

hope no one here interprets single initiator zoning as (just) "one hba and one sp port." thats "single initiator - single target zoning".

EMC does recommend single intiator zoning but they do not clarify whether it can be single target or multiple target zoned to single initiator.

I probably should have read up a bit before posting this but the initiator/target thing has me a little confused I guess. I hope we're doing it right.

Every time I zone for a new server to connect to my CX500, I add a zone that looks like this:


First of two identical switches
Server WWN Adapter 0
SP-A-0
SP-B-1

Second of two identical switches
Server WWN Adapter 1
SP-B-0
SP A-1

We make yet another zone for the server/jukebox drive relationship:

1st Switch
Server WWN Adapter 0
Tape Drive 0

And
2nd Switch
Server WWN Adapter 1
Tape Drive 1


What we've been avoiding is throwing them all together like this.

Server Adapter
SP-A
SP-B
Tape Drive

Somebody, somewhere told us to keep the zones tight. Like one server to one other thing so we have multiple zones containing the same server that connects to multiplel things.

i would rather disagree with this support tech.

considering following situation,
Zone1 -> hba1 - spa1 - spb1
Zone2 -> hba2-spa2- spb2

and then this situation
Zone1 -> hba1 - spa1
Zone2 -> hba1-spb1
Zone3 -> hba2 - spa2
Zone4 -> hba2-spb2

in both the cases, the hba1 sees spa1 and spb1...its completely transparent to host and hba and you cannot tell what kind of zoning you are using unless looked up on switch. so its upto you to decide which way you want to go.

as far as failover and tresspass is concerned, those problems would occur when using non-EMC failover software (i.e. powerpath).
Powerpath always tries to minimise the trespass when connected to clariion since the trespass penalty is high in terms of time. powerpath always tries to redirect io on owning sp through alternate HBAs (failover) and if thats unsuccessful, "trespass" is initiated.

other failover software like veritas may not do this if they are not aware of active-passive configuration like CLARiiON.

8< 8< 8< 8<

Hi guys,

I thought it maybe useful to define some of the processes and interactions that occur on a switch.

Firstly, when the HBA connects to the switch, it performs a fabric Login (FLOGI). During this login process, the switch assigns the HBA a FCID which is a unique address on the fabric.

The WWN of the HBA and its corresponding FCID is then registered with the fabric name server which holds similar entries for all other devices on the fabric.

This is where your zoning comes in - zoning is effectively a filter on the list of WWN's and FCID's in the nameserver - when our HBA asks the nameserver "who else" is on the fabric, the nameserver responds only with the FCID's of other devices that are "zoned" to the HBA that asked.

So now lets talk about initators...

Once the HBA has its list of FCID's from the name server, the HBA driver will attempt an N-Port login (PLOGI) to each of these addresses (targets) in the list to start communications - during this process, they will swap WWN info and determine (hopefully) that they can each talk SCSI.

The final step is when the HBA (initator) performs a "process" login to the target (PRLI) where they can then start SCSI communications with each other...

The reason for single initiator zoning is that it prevents the nameserver from returning the FCID of another HBA which in turn stops initators (HBA's) attempting to login to each other - this can cause connection issues with some HBA types...

OK, from the HBA's perspective, if you make a single zone with "HBA, SPA, SPB" or 2 zones with "HBA, SPA" and "HBA, SPB" there is no difference as the nameserver will return the FCID's of SPA and SPB in either case.

What is different is that with the single zone "HBA SPA SPB" configuration, the clariion SP's can see each other (if they bothered to ask)...

So in the case of our clariion using a single zone, it is possible (from a fabric perspective) for each of the SP's to talk to each other, I must say that I personally have not seen that to make any difference to a successful failover - but happy to be corrected though if someone has got some more info.

You will see that some devices act as both initators and targets at the same time - the Clariion SP's are examples of this if you have Mirrorview or SANCopy - the SP will behave as an initiator in those cases as it talks to the other array, but can also act as a target as per normal...

Hope that helps a little..

Glen.

That was an excellent contribution. I love forums for this reason. What you said makes a lot of sense and I'll be tucking that one away.

Thanks so much!

ppage

If you install SANcopy on the array, the clariion SP ports now become initiators and WILL log into each other if in the same zone - there is a restriction in our configuration guide about this.

glenk