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

Re: [Linux-cluster] Clustered LVM locking issues



Hello,


I might have misunderstood, but:

I am assuming one machine is exporting local storage via iSCSI.
Might it be easier to use LVM om the base storage device and take your
snapshot there?
(by exporting a LV as iSCSI target)
(keep inmind, this machine would be a SPOF for the entire cluster)


If you are using a Shared Storage Device and exporting the iSCSI target
using a cluster, this would be a problem, as this would also be a Clustered
LV.
In this case, advanced Shared Storage Devices offer internal LVM snapshots,
which do no interfere with the LUN. (HP MSA devices offer this)
It will allow you to chreate a LVM shapshot of your (iSCSI/SCSI/FC)LUN
without interfering with the 'O/S LVM layer', therefore allowing you to
continue exporting your device.


Regards,

Kit

-----Original Message-----
From: linux-cluster-bounces redhat com
[mailto:linux-cluster-bounces redhat com] On Behalf Of Paolo Smiraglia
Sent: maandag 13 december 2010 19:18
To: linux-cluster redhat com
Subject: [Linux-cluster] Clustered LVM locking issues

Hi to every ones....

We have configured a shared storage between some nodes with iSCSI, and we
want to use LVM across them.

Our access model is based on syncronized commands remotely executed by the
master node with ssh (authenticated by public key without password).

The master node is the one that can create/remove logical volumes which are
a snapshot of a "base" logical voume.

Other nodes exclusively access logical volumes created by master node.

In order to do so, we have installed in all nodes

  * RedHat Enterprise 6 beta2
  * Cluster Suite
  * CLVMD

Then we have configured LVM with

   locking_type=3 (lvm.conf)

and fist attept we marked the volume group as "clustered".

Unfortunately we got an error message that is saying snapshot is not
supoprted for clustered volume groups. In order to overcome this issue we
used a workaround that conists in clearing the clustered flag in the volume
group. However, this caused some concurrency problems that prevent a logical
volume to be removed even if not used.

Do you have a solution for this problem? Thanks in advance for replies...


--
PAOLO SMIRAGLIA
http://portale.isf.polito.it/paolo-smiraglia

--
Linux-cluster mailing list
Linux-cluster redhat com
https://www.redhat.com/mailman/listinfo/linux-cluster


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