Matt Domsch ha scritto:
On Thu, May 19, 2005 at 02:11:43PM +0000, pg@xxxxxxxxxx wrote:As I'm not so expert in HA clusters I got an hw ans sw solution "suggested" by DELL and my hw/sw configuration, that is working quiet well since a couple of years. May be i've been luky.
I exerienced the following error on a RedHat cluster configuration with Dell hardware (Perc 3/DC controller and PowerVault 220 disk array). When the error occurs the cluster manager shutdown the cluster node, but the filesystem is corruped and the other node cannot mount it until a manual fsck.
Any idea?
The firmware on the PERC 3/DC is not multi-initiator cluster-capable under Linux. For this reason, neither Dell nor Red Hat recommend trying to create a HA shared storage cluster with this configuration. Even with write cache disabled, the lock sectors used by the cluster manager don't manage to stay coherent.
I understand that newest versions of the Red Hat Cluster Suite may no longer use lock sectors on the disk as the I/O fencing mechanism, which may then enable such configurations. But neither Dell nor Red Hat have done any testing with the hardware config you've got.
The price is right, until you actually need your data to be highly available and it crashes.
Thanks, Matt
As racommended I don't mount the same filesystem on both nodes of the cluster at the same time; every service has its own filesystem and a service is active on a single node.
The uniqe shared partition is the quorum, that is on a RAID-1 volume. The other partitions are on a single RAID-5 volume: do you think that to make a seaprate volume for each partition could help?
Thanks
- : send the line "unsubscribe linux-scsi" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
