Originally posted by: Landao

Hello guys

I have IBM rs6000_1 and rs6000_1fb hosts machines and a shared disk array between this machines.
Two volume group: vgdata_sh and vgcode_sh are in diskarray.
This means that one machine owns the VGs and hence writes data to them, but should that machine fail, then another machine would then take control of the disks, and has immediate access to all the data that the failed machine had.

In currently vgdata_sh and vgcode_sh VGs are used by rs6000_1 machine. But when I turn rs6000_1 machine off I couldn't take control this volume groups from rs6000_1fb machine by means of following commands:

importvg -y vgdata_sh hdisk4
importvg -y vgcode_sh hdisk5
varyonvg vgdata_sh
varyonvg vgcode_sh

Who any one let me know advice how can I take control this VGs to rs6000_1fb ?

Thanks in advance
#AIX-Forum

Originally posted by: alethad

I would suggest modeling your systems after an HACMP cluster configuration. And then when you make changes use what we call the Lazy update method to get the definition changes on the inactive node.
Do you have the disks & volume groups seen by both systems with both systems up & running? You have to get this done first.
If the disks are seen by both systems then you need to get the volume groups varied off of the primary system. Then import them onto the secondary system and vary them on. Mount up the filesystems too. Make sure everything is working order. Then unmount the filesystems and vary off the secondary systems. Then vary them back onto the primiary system.
<hr />
Once you've gotten the volume groups done whenever you make ANY change to these volume groups you will need to re-import them on the secondary system. Here's what I do:
Primary node:
varyonvg -bu vg_name

Secondary node:
importvg -L vg_name hdisk#

Primary node:
varyonvg vg_name
<hr />
I'm sure someone may have a better suggestion but you're doing the same principle as I do with HACMP.
If you need more detail I would be glad to send it to you.

Good luck.
#AIX-Forum

Originally posted by: Landao

Hello, alethad, first of all I thank you to pay attention to my problem.

Yes we have not HACMP cluster configuration but the disks & volume groups seen by both systems with alternate by doing the following commands manually. Thatis, I do:
On primary node:
1. varyoffvg vg_name
2. exportvg vg_name

On secondary node:
1. importvg -y vg_name hdisk#
2. varyonvg vg_name

After issuing this commands I am able to take control the valume group and have access to all the data in secondary machine that the primary machine had.

But problem that in the event of a disaster situation thatis primary node fail I could not switchover the volume group to the secondary machine. (without issuing above commands on the primary node.)

Best regards
#AIX-Forum

Originally posted by: alethad

You shouldn't need to export the volume group on the primary node. You probably need to look at setting your volume groups with the -C option. It marks the vg as being enhanced-concurrent without really being a concurrent vg. Read up on this. You can turn this on dynamically but you can't turn it off without rebuilding the vg's. If you can create a vg to test it that would be good.

What I would like you to see is when you run an lspv on both nodes you see all the disks with their volume groups. Only node1 would have control. You should be able to on the primary node just get the vg's vary'd off(no export) and turn right around and vary them on on the secondary. If node1 one is down they vg's are basically already vary'd off. So all you need to do on node2 is varyon and mount fs's. If you can do this with both systems up you should be able to do it with node1 is down.

I may not be explaining this very well. But I've done this many times but with the -C set. On my nodes I only have to unmount fs's & varyoff my vg's on node1 and then I varyon the vg's & mount the fs's. I can't think of anything else right now that I do to config this.
Maybe someone can jump in to clarify if I'm missing something.
Let me know how it goes for you.
#AIX-Forum

Originally posted by: jvk

It seems to me that you hit a reservation conflict on a second node when your VG is not properly closed down on primary node (means regular varyoffvg).
Without HACMP you need some tool to break reservations and AIX has no such tool. Look on the storage site for this. What kind of storage do you use?
#AIX-Forum

Originally posted by: Landao

Our machines are IBM p5 Model 520 and OS is AIX5.3

We use IBM DS4700 disk storage and properties are :
2x 4GBIT RAID controllers (2GB Cache, up to 4GB)
4x 4GBIT Host Ports (2x at each controller), up to 8
4x 4GBIT FC External Drive Interfaces (2x at each controller)
4x Short Wave SFP GBIC (LC) w/ LC/LC cables
6x 73GB/15K FC4 Hot-Swap Hard Disk Drives (up to 112)
2 Storage partitions (up to 64)
2x Redundant Power Supplies & Fans
Total Raw Storage Capacity ~ 430GB (up to 33.6TB )
Transfer Throughput Performance: ~ 1600 MB/sec
Transfer I/O Performance: 125,000 IOs/sec
Support RAID levels 0, 1, 3, 5, 10
AIX Host Connect License
#AIX-Forum

Originally posted by: jvk

Ok, you'll need a DS4700 expert to confirm that you can do something like this:

1) check with the vdShow command which controller owns the volumes in question. This is the command from the shell of controller.
2) connect to the correct controller with a terminal (i.e. your PC with a serial connection and putty for exmple) and run a couple of commands from cmd line:
3) hlmTestLunShow ( volume # )
hlmTestRelease ( host id, volume # )

The first command will print out several pieces of information, one of which is the host id of the reservation holder. (Note: host id=reserve id in above hlmTestLunShow output.) The second command will release the reservation.

Example for LUN 3

Controller A:
-> hlmTestLunShow 3
LunNumber:0x3 LunInfo :0x40ff9664 State:0x0 QuiescenceFlag:0x0
Owner:0x0 IsReady :0x1
reserveId:0xffff resv3rdId:0xffff
value = 132 = 0x84

If the reserveId is 0xffff then there is no reservation on that volume (or you are on the wrong controller).

Controller B:
-> hlmTestLunShow 3
LunNumber:0x3 LunInfo :0x40ff9664 State:0x0 QuiescenceFlag:0x0
Owner:0x1 IsReady :0x1
reserveId:0xa resv3rdId:0xffff
value = 129 = 0x81

-> hlmTestRelease 0xa,3
LunNumber:0x3 LunInfo :0x40ff9664 State:0x0 QuiescenceFlag:0x0
Owner:0x1 IsReady :0x1
reserveId:0xffff resv3rdId:0xffff
value = 132 = 0x84

Now LUN 3 should be accessible for the second node.

Btw, above procedure is not supported by IBM, they'll request you use HACMP which has its own tools to break reservations during the takeover.
#AIX-Forum

Originally posted by: SystemAdmin

test post
#AIX-Forum