It was determined that a node reboot would force an actual rescan of the disk controller and recognize the disk as available for addition. Again unfortunately, though, once another disk in the node failed the empty bay issue would return. This would again require a "draining" of connections and a reboot of the node, which really impacts the availability and administrative effort required by the customer.
It was determined that the actual fix was to ensure the firmware was updated to the latest version. Apparently there were issues with the LSI disk controller not actually performing a rescan when requested by a software level command. The latest firmware ensures that the software command called by OneFS actually results in an LSI controller rescan, therefore resulting in a disk addition without taking the node offline.
Of course this does mean that a reboot per node was required to perform the firmware upgrade, but that is a lot better then a reboot every time a disk fails.
BlueSteelGT
17 Posts
0
July 16th, 2013 19:00
It was determined that a node reboot would force an actual rescan of the disk controller and recognize the disk as available for addition. Again unfortunately, though, once another disk in the node failed the empty bay issue would return. This would again require a "draining" of connections and a reboot of the node, which really impacts the availability and administrative effort required by the customer.
It was determined that the actual fix was to ensure the firmware was updated to the latest version. Apparently there were issues with the LSI disk controller not actually performing a rescan when requested by a software level command. The latest firmware ensures that the software command called by OneFS actually results in an LSI controller rescan, therefore resulting in a disk addition without taking the node offline.
Of course this does mean that a reboot per node was required to perform the firmware upgrade, but that is a lot better then a reboot every time a disk fails.