VA7400 and duplicate LUNs appearing in Sam

[Michael Gretton]

I thought there was a "storage forum" but I cannot find it in the category pull-down. WE have had a VA7400 sitting around here for a while with no time to set it up. I am now just getting the opportunity to do so. I have created a few test LUNs and I have given my server the proper permissions to configure and see the LUN. All is well and I go to SAM on my server and I see the LUN. However, I see it TWICE. This is for all LUNs that the box has access to. What gives? It is strange but I can manipulate both LUNs as though they are two separate LUNs even though I know that they aren't!

I have two controllers on the back of the VA7400 going into one brocade switch. This is the same switch by the way that the servers are connected to. This was installed per HP reps suggestions.

I have read the manuals to secure manager, the VA7400 and etc. and I am not getting anywhere. I have also emailed my HP rep who helped me get started and unfortunately he is no longer there. I can place a trouble ticket in but I would rather try all options first.

Points will be given accordingly. Please help. Thank you,

[Ted Ellis_2]

sounds like your server is seeing each LUN across both of the controllers. The server will see both controllers and both controllers are seeing all of your LUNs. It is a good thing, as you can configure your server with some hardware failure protection.... if one controller fails then the server communicates across the alternate. Do you have dual fibre adapters from the server to the switch? You will want to create the volume groups using the LUNs so they have alternate links enabled... basic idea is that you vgcreate using one of the LUNs per normal. You then use vgextend to add the other path to the same LUN.... it will automatically register the second instance as the alternate link.

You can confirm that this is the case by noticing a difference in the hardware path for the duplicate entries.... we specifically use dual adapters on the server so we can also survive a fibre adapter failure

Ted

[Michael Gretton]

Ted:

I only have on HBA per server. I suppose I need to read up on alternate paths. I am confused. once I configure the LUN the "second" lun appears as unused. What is to prevent me in the future, forgetting that I already used that lun, reconfiguring it and losing information?

Mike

[Ted Ellis_2 · Accepted solution]

as long as you don't do something like pvcreate -f, the LUN will report back that it is use in another volume group.... so you should be ok. Do you see two different hardware paths for the two instances of the same LUN. Most likely have all numbers in the path the same accept for one.

run ioscan -funC disk and look for the LUNs and their device files. You can confirm which one is used in your volume group by running vgdisplay -v vgxx... it will end with a report on the physical devices and give you the device file. Correlate that to the ioscan to determine which one is the primary. Now to configure the alterante:

locate the device file from the ioscan that correlates to the unused disk (LUN);

vgextend /dev/vgxx /dev/dsk/cxtxdx

that will register the alternate link, as it will hit the LUN and see that it is already configured as the primary disk.

hope that helps

Ted

[Ted Ellis_2]

I think you may also get some good mileage by reposting this to the Storage forum... then you can get some attention from people with specific VA knowledge to assist with configuration, etc.

Ted

[David Bell_1]

Michael,

You can also put "Usage" in under SAM. This will show a "visual representation" that the volume is used. So under the controller that shows the LUN as unused, you can put a comment "CTL_1_Configured" or something of that nature. I'm not sure of the limitations in that field. You can do a "mark disk as used" under the Actions Menu.

You may also want to verify that you have the latest WSIO SCSI and SAM/Obam patches. This may also effect the way in which you see LUN usage under SAM.

HTH,

Dave

[Tom Danzig]

Are linked up to the VA via a SAN (i.e. Brocade switch)? If so, you're seeing the path to each LUN via each controller attached to the SAN; not from an additional controller.

[Tom Danzig]

My last post should have read:

Are linked up to the VA via a SAN (i.e. Brocade switch)? If so, you're seeing the path to each LUN via each controller attached to the SAN; not from an additional HBA.

[Jeff Schussele]

Hi Michael,

Have you set up any VGs on any LUNs yet?

If so do a
vgdisplay -v /dev/vg_name
and check the output for "alternate links"
If you see them then you DO have 2 separate paths to the same LUNs.

Rgds,
Jeff

[Ashwani Kashyap]

SO there is only one HBA per server , and the two controllers from VA7400 goes to the same brocade as your HBA .

Is there zoning set on the switch . If yes , then I think that you might have zoned the HBA twice , to each of the controllers on the VA7400 who are seeing the same set of disks .

[harry d brown jr]

This member has assigned points to 15 of 184 responses to his/her questions.

And you say:
Points will be given accordingly. Please help. Thank you,

Really? Funny, what about these posts: http://forums.itrc.hp.com/cm/TopSolutions/1,,CA400109!1!questions,00.html
??


And yes there is a storage forum section: http://forums.itrc.hp.com/cm/FamilyHome/1,,120,00.html


live free or die
harry

[Michael Gretton]

Thank you all for your responses. Very helpful. I am going to go through all the suggestions. This morning I did confirm that I am getting separate hardware paths from each LUN. One number off. Thanks a bunch Ted et al. One LUN is 1/0/0/0/.1.18.0.0.0.1 and the other has a 19 where the 18 is.

Harry:

You keep bringing this up and bad mouthing me in every post I bring up. Not everyone lives on this board and understood the point system from the get-go. Many people assume that this board is strictly a way to communicate via HP and is paid through support. If I pay for this service (which I pay annually 65K to HP), assigning points is my last priority during the day. However, I know this to be different and that the people on this board help me out on their own time. Further, the point system is a good incentive to get people to participate.

The reason I said that I would assign points accordingly was to make it clear that I do assign points. I may not always have done this but I do now. I have understood the error of my ways. Further, I will fix the point assignment on my previous questions when I get to them. I would appreciate it if you would stop bringing this up everytime I post a message. I apologize if cleaning up my old message point assignment is not being done quickly enough for you.

Again, thanks everyone for helping out. I will look for this storage forum. I do not get it in my drop down box, however.

[Eugeny Brychkov]

Mike, as you configure paths for your LUNs created within VA please know that there's a perfomance rule: if you create LUN within Redundancy Group 1 (RG1, see armdsp -a output) then primary path should lie through controller 1 (C1) of VA, if you create LUN within RG2 then primary path should lie through C2.
For VA7100, where only one RG1 exists, all primaries should lie through C1.
In addition, when configuring VA please check resilency setting: the best is "normal perfomance" when maps are written to map disks every 4 seconds. If you choose "high perfomance" maps are written only on correct VA shutdown
Good luck!
Eugeny

[Michael Gretton]

Eugeny:

Thank you for that reminder. When my HP rep came in to help me, he talked about that. However, it is hard for all that to sink in on your first day with a a SAN and absolutely no experience. But it is coming back to me now. I am going through Ted's comments and when I get things configured correctly, I will let you all know how I did.

Mike

[Michael Gretton]

Eugeny:

What is the best way to know which path is C1 and which path is C2?

Mike

[Eugeny Brychkov]

Mike, to get know where C1 path and where C2 path you should disassemble harware path. If you have quickloop check AL-PAs of controllers, if fabric then you can identify path by switch's port #.
You can have some fun and visualize your results unplugging FC cable of one controller and look to ioscan which path will have "NO_HW" state... but it's good to do while you're in testing phase.
Eugeny

[Eugeny Brychkov]

"HP Fibre Channel Fabric Migration Guide":
http://docs.hp.com/hpux/onlinedocs/J2635-90014/J2635-90014.html
"Installing and Administering HP Fibre Channel/9000 ":
http://docs.hp.com/hpux/onlinedocs/J2390-90010/J2390-90010.html
These all links I was found for now. Try searching www.hp.com and you may find more info you need.
Eugeny

[Tom Danzig]

"What is the best way to know which path is C1 and which path is C2?"

I would do a dd on each path and see which controller "lights up". Also, the access time from the performance path will be significantly lower than the non-perf path.

Do the following on each of the devices (where cX will be different and cYtZ will be the same):

timex dd if=/dev/rdsk/cXtYdZ bs=64k count=1000 of=/dev/null

[Tom Danzig]

Damn ... wht can't I spot the types before posting!?

Anyway, above should read:

(where cX will be different and tYdZ will be the same):


(No points for this post please)

[David Bell_1]

Michael,

I didn't know if you had this so I thought I would share it. This pdf shows the definitions of what you're seeing in the ioscan. Further, in order to get the interfaces and associated "N-Port ID's", try using . You could also get some useful information from where 'X' is the number of the HBA interface device driver, typically td0 and td1 if you have two HBA's. See the following URL where I have posted the PDF previously. Look at the 'fabric' based pdf not 'private loop'.

http://forums.itrc.hp.com/cm/QuestionAnswer/1,,0xd32136e69499d611abdb0090277a778c,00.html

HTH,

Dave

[Michael Gretton]

I know it has been a while since I post this question. I just wanted to let everyone who helped know that I got things configured further thanks to your help. Ted, David, Tom, Eugeny and others - thank you. David, the link you provided really helped.