[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]

Re: [Linux-cluster] lvm locking type 2, still used?

With that said, the latest lvm2-cluster-2.02.21-7.el4 installs /etc/lvm/lvm.conf.rpmnew, with new locking_type definition(s).  It leaves the previous lvm.conf alone, which we did, and that is set to #2.  I am not clear on the use of Type 3 -- does that mean there is a static version, for like a ramdisk?  Which is preferable settings for a 'regular' cluster LVM with GFS?  Please advise, thanks.

    # Type of locking to use. Defaults to local file-based locking (1).
    # Turn locking off by setting to 0 (dangerous: risks metadata corruption
    # if LVM2 commands get run concurrently).
    # Type 2 uses the external shared library locking_library.
    # Type 3 uses built-in clustered locking.
    locking_type = 1
    fallback_to_clustered_locking = 1
    fallback_to_local_locking = 1
    locking_dir = "/var/lock/lvm"

On Wed, 2007-05-30 at 15:31 +0100, Patrick Caulfield wrote:
isplist logicore net wrote:
> I am adding nodes and just noticed something interesting. When I started with 
> GFS/Clustering, one of the requirements which I made a note of was having to 
> modify the /etc/lvm/lvm.conf in the following manner for my RHEL4 machines;
> locking_type = 2
> and
> locking_library = "/usr/lib/liblvm2clusterlock.so"
> I just noticed that over time, this file was changed back to what it used to 
> be by default which is;
> locking_type = 1 
> and 
> locking_library path has been removed. 
> This is not something I've done, it's happened over time, using yum to keep 
> things up to date. Yet, the cluster works just fine and all volumes can be 
> seen. Did I miss something about not needing to do this anymore?

The locking is only really needed when you change LV metadata (lvcreate,
lvextend etc). so if all you are doing is activating existing LVs then the
missing locking will make little difference.

If you make any changes to LVs on a mode with the locking_type set incorrectly
then those changes will not be visible on the other nodes. However, lvm2 is
smart enough not to let you change clustered VGs when clvmd isn't available so
it's unlikely to have caused any harm. you would have noticed something was
wrong the first time you needed to change anything/

Robert Hurst, Sr. Caché Administrator
Beth Israel Deaconess Medical Center
1135 Tremont Street, REN-7
Boston, Massachusetts   02120-2140
617-754-8754 ∙ Fax: 617-754-8730 ∙ Cell: 401-787-3154
Any technology distinguishable from magic is insufficiently advanced.

Attachment: smime.p7s
Description: S/MIME cryptographic signature

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]