Start a Conversation

This post is more than 5 years old

Solved!

Go to Solution

1928

February 12th, 2008 08:00

VCM disk showing up on SAN attached host

I just noticed this and seem strange to me: some hosts that SAN attached
see the VCM disk, even the disk is not masked on the directer port.

Enviroroment: EMC DMX 1000 ( Mcode 5670 ) and DMX3000 ( Mcode 5671).
hostA have solution enabler installed. while hostB did not.

I have two questions:
1. on hostA
hostA>#symdev -sid xxxx list | grep VCM
0000 c2t5006048ACAFE1E6Cd0* 13C:1 16A:D0 2-Way Mir N/Grp'd VCM WD 45

so device "0000" which is a VCM disk see hostA see as "c2t5006048ACAFE1E6Cd0" .

hostA>#echo |format | grep -i e6cd0
c2t5006048ACAFE1E6Cd0: configured with capacity of 44.06MB
2. c2t5006048ACAFE1E6Cd0

hostA>#symmaskdb -sid xxxx -dev 0000 list assignment

No device masking database records could be found for the specified input parameters

so why does lun "0000" showing up at all on hostA? does this have
anything to do with solution enalber?

2. on hostB, that DOES NOT have solutions enabler,
it also see the VCM disk ,and even worse, one of my SA's trying to
format and label the disk and unable to do that ( lucky we :- ).

so the question is: why can't format and label the VCM disk ?
is that because the state is WD ( write disabled )? and why
the VCM even showing up on this host?

Thanks.
JasonJ

2.8K Posts

February 12th, 2008 14:00

It's difficult to give a general rule that always work since the configuration of the box can change a lot from customer to customer. But generally speaking, the default behaviour is that each and every host that is attached to the SAN should see the VCM. There are exceptions .. A lot of exceptions. You can have lun offset, you can have ports where the VCM isn't mapped even if it's mapped on the other port of the same processor .. It's also possible to "hide" the VCM (just like any other volume not masked). It's very complex to give a rule as I stated before :D

The VCM is WD just to avoid SysAdmins to mess your vcmdb :-)

2.8K Posts

February 12th, 2008 12:00

When you need to connect a brand new host to the storage, you need at first to create the zones on the switches. But how can you know if the zones are working if you didn't yet mask volumes to the host ?? That's why the VCM exists :D
The VCM is a very special volume that is always available even if no explicit masking exists. It's enough to map the volume with the VCM flag and all the hosts zoned to the FAs will see this very special volume.

In good old DMX (DMX1 and DMX2) the VCM is still a volume (24, 48 or 96 cyls big) whith another very important role: it contains the symmask database itself :D .. Since it contains such an important database, it's kept Write Disabled just to be sure that nobody will tamper it.

In new DMX (DMX3 and DMX4) the database is in another area (SFS - Symmetrix File System) of the storage and you don't really need a volume where to put the database. But you still need a volume always visible. So the flag VCM still exists and is usually assigned to a GateKeeper that's mapped to all the ports of the DMX.

Just to return in topic .. It's normal to see the VCM device even if no masking exists. Solution Enabler is our software that allows you to manipulate VCM and masking .. But you don't need S.E. to "see" the VCM :D .. From your message I can tell you that you have a DMX1 or DMX2 (models ranges from 800 to 3000) since the VCM is still WD (while with newer boxes it's RW enabled since it's a GK). Now you know why it's WD ;-)

Cheers !!

50 Posts

February 12th, 2008 12:00

Thanks for answering my question. my concern is why the VCM in DMX1000/3000
is even need to be visiable on the host side?

my reasoning is that VCM database is checked by the FA directors to see which host have accesx to which LUNx on that FA, so it is fine to allow FA have accees to VCM disk, but why the host need access it anyway?

Thanks.
JasonJ

50 Posts

February 12th, 2008 13:00

Thanks. Stefano. so you seem to suggest VCM disk WOULD and SHOULD to be seen by every hosts that SAN attached and zones into the Symmetrx DMX?

what is confusing is at least in our enviroment that contains DMX1000( Mcode=5670) and DMX3000(Mcode=5671) , some hosts see the VCM disk,
some host don't. ( as checked via format ). any suggestions about why? would this maybe depending on solaris versions and/or HBA drivers ?

adding to the confusion is when the host does see the VCM disk, user or SA's will try to use this disk when they shouldn't.

Thanks for clarifying for me.
JasonJ

2.8K Posts

February 12th, 2008 13:00

JasonJ .. when you zone your host, the VCM is the only volume you will see. And when you see the VCM you can say "OK the zoning is good, we can go on and mask some useful devices". Think to the VCM just like the way to verify that the zoning is good. It's also true that each and every FA needs the VCM to lookup the database, but however each and every host will see the VCM just as a proof that the zoning is good.

Message was edited by:
Stefano Del Corno

50 Posts

February 13th, 2008 08:00

on VCM Ddisk ... "Since it contains such an important database, it's kept Write Disabled just to be sure that nobody will tamper it."

to clarify on the "Write Disabled" flag: does it means:
Write disabled for host, but still RW for DMX and FA directors?

Thanks.

2.8K Posts

February 13th, 2008 10:00

Try to run "symdev show" against your vcmdb .. Usually a device is write disabled only on the frontend .... However don't pay too much attention to the status (write disable) of the vcmdb .. put more attention to its role and take great care not to mess with it via symmask commands :D

76 Posts

February 27th, 2008 22:00

Can anyone tell me how to delete the VCM entry from the host.?

We have a AIX 5.3 OS and we decommisoned DMX luns from this server.Even I removed the zoning.

The User DOESNT WANT TO SEE the VCM 0000 lun when he runs the inq command.

Can you tell me how to disappear this lun ?

2.8K Posts

February 27th, 2008 23:00

The Host still have Zones but not zoned to DMX but ti HDS ! ;)

Do u want me to remove the fibre cables ! lol ..


Since your host is a betrayer, it's a good idea to unplug all the fibres ;-)

What might be wrong here ?


If you really can see the VCM from our DMX, that means you still have at least a zone that goes from one of your HBA to an FA .. that's it. :D

76 Posts

February 27th, 2008 23:00

The Host still have Zones but not zoned to DMX but ti HDS ! ;)

Do u want me to remove the fibre cables ! lol ..
What might be wrong here ?

2.8K Posts

February 27th, 2008 23:00

If you removed the zones the host simply shouldn't see ANY device from the DMX :D

The VCM itself can not be hidden since you can't either "grant" or remove grant to an host. The VCM can be seen as long as the zoning is right. If you still can see the VCM (and you can show with the good old "dd" command its content) it's likely that the host still have zones ;-)

76 Posts

February 28th, 2008 00:00

Trust me !!! THe host is not connected to DMX !!! :D

Anyways,I have asked the user to ignore this disk ;)

2.8K Posts

February 28th, 2008 00:00

Trust me .. if the host can REALLY see the VCM, you still have something in between that allows the host to see the VCM ;-)

Did you reboot the host after removing the zones going to the DMX ??

Usually switches do implement the so called "soft zoning" .. Google a little to find better details, however to keep things short, it's a zoning implemented in the Name Service that the switches do always offer. You simply define a list of wwns that can talk one to each other (a zone). When the HBA logs into the fabric, it will ask the Name Service "Who can I talk to, today?" And the NS will answer supplying the list you configured. But if you later change the zoning (in the NS) and the host doesn't ask again "Who can I talk to?" nothing will stop it from sending frames to previously allowed destinations.

Since you are absolutly sure that there is no active zone going from your host to the DMX and you are also absolutly sure that you can still see the VCM, the only idea that comes to my mind is that you are using the so called "soft zoning" (that's the default on almost every switch) and your host didn't log out and in again on the fabric.

If possible, try to reboot the host .. and hope it does work :D

76 Posts

February 28th, 2008 01:00

Yes we always soft zone .

And hey I asked this query after rebooting the host ! :D

2.8K Posts

February 28th, 2008 02:00

If you still have Solution Enabler on the host, please run the following commands and show the output .. You wake up my curiosity !! :D

1) syminq | grep -i emc
2) symmask list hba
3) sympd list

ThX :D
No Events found!

Top