&KDSWHU&4XLFN6HWXS
After the completion of this process, the disk array's state changes into
ready
again.
A few words on the replacement of a defective hard disk of a disk array.
If a hard disk belonging to a disk array for which no Hot Fix drive had been assigned should
fail, you should replace this defective hard disk with a new one
as soon as possible
. Always be
aware of the fact that this disk array does not have any redundancy until the defective hard
disk has been substituted. This means that if another hard disk should fail while the disk
array is without redundancy, all data is irretrievably lost. RAIDYNE offers two possibilities of
replacing a failed drive of an array for which no Hot Fix drive has been designated:
1. Replacement with GDTSETUP (we have just demonstrated this method)
2. Replacement by using the
Hot Plug function of GDTMON
The Hot Plug method allows you to replace a defective drive while the disk array continues
to work and without having to load GDTSETUP. When this method is used, the GDT SCSI
bus to which the defective drive is connected, is temporarily halted (that is, for the time
necessary for replacement), thus enabling you to disconnect the defective drive from the
SCSI bus without any risk. After the replacement, the SCSI channel halt is lifted again and
RAIDYNE automatically begins to rebuild the new drive. The halting and release of the GDT
SCSI channel is controlled by GDTMON, which is available for most operating systems sup-
ported by the ICP Controller. The above mentioned halt of the SCSI bus serves to avoid that
interferences (spikes and glitches) which inevitably occur on the SCSI bus when the defec-
tive drive is disconnected, impair the functioning of still intact drives on this SCSI channel.
However, this implies that none of the SCSI devices of the halted SCSI channel can be ac-
cessed during the time the defective drive is being exchanged. If all drives forming the disk
array are connected to one single SCSI channel the entire disk array cannot be accessed
during the time of replacement. Therefore, it is evident that the ICP Controller should have
as many SCSI channels as possible and that all SCSI devices should be distributed equally
to the available channels in order to avoid that the disk array or other SCSI devices cannot
