LockLun vs Lock disk

[boomerang]

Hello Admins,

Please help me to understand the difference between Lock Lun & Lock disk

[Patrick Wallek · Accepted solution]

Different terms for the same thing and they perform the same function.

A "lock lun" is referring to a LUN presented from some sort of SAN storage that

A "lock disk" could be a physical disk drive that all servers in the cluster have access to.

[Matti_Kurkela]

For Serviceguard, there is not much difference.

When exactly one-half of the cluster has failed, or one half of the cluster has lost connection with the other half, Serviceguard uses a "cluster lock" to decide which half of the cluster is allowed to continue; the other half must stop, since it cannot coordinate its actions with the other half. When a cluster has only 2 nodes, the cluster lock is very important; in clusters of 4 or more nodes, it may not be necessary.

The cluster lock can currently be implemented in one of three ways:
a) using a Serviceguard Quorum Server
b) using an external physical disk (lock disk)
c) using a LUN on a SAN storage system (lock LUN)

The lock disk and lock LUN are configured exactly the same way: the only difference is the hardware used to implement it.

The cluster lock must be accessible by all nodes, even if one of the nodes is powered down or totally destroyed. On older hardware, a common solution for 2-node clusters was to set up an external HVD or LVD SCSI disk box and connect it to both nodes using SCSI cables with special in-line terminators. This requires that the nodes are physically close to each other to conform with SCSI cable length limits: setting each node to a separate server room might be difficult or impossible. (With modern SAS disks, this kind of set-up may no longer be possible.)

On modern hardware, a Serviceguard cluster is often set up using an external FibreChannel SAN storage system. FibreChannel has very generous limits on cable lengths and the number of disks and nodes you can use. On enterprise-class SAN hardware, the disks are "virtualized" by the SAN hardware to allow the SAN administrator more freedom in implementing various RAID levels and slicing the disk capacity as necessary. As a result, the servers plugged into a SAN will see configurable storage units that behave like disks, but aren't actual physical disks. They are "logical units" or LUNs. For HP-UX and Serviceguard, such a LUN is essentially the same as an actual physical SCSI-attached disk.

Older Serviceguard documentation tends to use the term "lock disk", while newer documents may prefer "lock LUN". They are interchangeable; the documentation just reflects what was the most common way to implement a disk-based Serviceguard cluster lock at the time the documentation was written.

MK

[Duncan Edmonstone]

I'm going to respectfully diagree with some of my fellow posters here... I think there are more differences between a lock disk and a lock LUN than they indicate:

- A Lock disk has to be part of LVM - and therefore part of a volume group in the cluster. (it can still be used for application data though)
- A Lock LUN is _not_ part of LVM (in fact it can't be part of LVM, VxVM, ASM etc.), it is just a physical block device (/dev/dsk/cXtYdZ or /dev/disk/diskNN) - it can also be very small (100KB is generally sufficient). You cannot store application data on a lock LUN
- Lock disks can operate in single and dual configurations (dual cluster locks generally being used in stretch clusters). Lock LUNs can only operate in single cluster lock configurations.
- Lock disks need to be backed up (via vgcfgbackup after the lock has been written). Lock LUNs do not.

HTH

Duncan

[Matti_Kurkela]

Thanks for the correction, Duncan.

I had to re-read the chapter about cluster locks in a fairly recent "Managing Serviceguard" manual, and it seems you're correct.

Looks like the lock LUN functionality was added in Serviceguard A.11.18; I seem to have missed this new feature.

Time for me to read the latest release of "Managing Serviceguard" to refresh my knowledge, I guess. Live and learn...

MK

[Emil Velez]

IF you have a volume group being used by a package then use a lock disk. WHy waste a dedicated disk.

If you have a cluster where you are using vxvm only and vxvm disk groups or CFS filesysetms and you have a 2 node cluster then you can use a locklun.

IF you have a 4 node cluster you do not need a lockdisk or locklun. Lock disks, locklun or quorum server are only absolutely necessary with 2 node clusters.


In the US we have a 4 hour online seminar that outlines the changes with the serviceguard product that people can attend that may help.

http://www.hp.com/cgi-bin/education/schedule.cgi?cnum=u5458aae