Hi, I have a server with a MegaRAID SATA 300-8X. The controller has 5*1TB drives attached. Previously I had 5*250GB drives on the controller all configured as a single logical RAID5 drive. But the controller only supports up to 2TB logical drives so after upgrading I had to configure 3 logical drives (sda, sdb, sdc) in order to make use of all the space. These are in turn configured into logical volumes with LVM. However, the performance seems to suffer very badly when accessing several of the logical drives concurrently. I'm guessing that this is because the IO scheduler maintains separate queues for each of the drives and assumes they can be accessed in parallel, causing lots of inefficient seeks across the array, degrading performance (and potentially wearing out the disks faster) int he process. Switching the IO scheduler to noop seems to help a little but by no means cures the issue. I wonder if it would be possible/desirable to somehow force all the logical drives to share the same I/O queue. Has this been done before for any other drivers? And is it even the right approach to resolve this issue? Although don't have much experience with SCSI drivers I would be happy to try to work on a solution if someone with knowledge of the LSI driver could offer me guidance. I'm not sure where it should be done, either in the MegaRAID driver itself, by exposing the logical drives as one joint SCSI drive to the kernel - or a more generic solution at a higher level by somehow joining the drives with a kind of master/slave setting using the /sys interface (e.g.: echo sda > /sys/block/sdb/master, telling the kernel that sdb is in fact a part of the same disk as sda and thus should share it's I/O queue etc.). I would welcome any insights in this matter. Best regards, Thomas -- To unsubscribe from this list: 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
