While not as simple as its automated hardware counterparts, software RAID under Red Hat Linux provides some level of redundancy. While setting up the arrays is somewhat of a hassle, rebuilding a member disk after a failure is a simple one-line command.
After coming into some recently purchased Itanium-based servers, it became apparent during the setup that hardware RAID controllers weren't included in the purchase price. Until hardware RAID could be ordered, an attempt to use the software RAID under Red Hat EL was initiated. Note that this software RAID is native to EL, as opposed to setting up evms. Setting up the array was simple enough in theory, although a bit of a pain using the quasi-graphical text interface; it took about 15-20 minutes to create several partitions on each of three disks, then create the software RAID devices using those paritions. Much of the time was spent navigating the infernal menus using the Tab key, and forcing the installer not to auto-assign the partitions to disk.
Once up, formatted, and running... the only thing left to do was to fail a disk and see what happens. So a disk was pulled from the running system... and the system kept running, making a one line entry in the syslog about the failed device. Now some amount of time passed before finding the appropriate manual page for checking the status of the arrays, and more importantly rebuilding the array that was just ruined on purpose. Hence the purpose of this entry, to make the command easier to locate in the future.
Warning: The following commands were issued on Red Hat Enterprise Linux Advanced Server 3.0 Update 4 on ia64 hardware, with minimal data to be lost. Not that it should matter too much, but just in case your mileage varies... don't say I didn't tell you so if it doesn't work and your array crashes.
The server reported that the /dev/sda1 parition was damaged when the disk was re-inserted. That partition was part of the /dev/md0 array. That in mind, the following command started the restore process.
# raidhotadd /dev/md0 /dev/sda1
During the restoration (or any other time for that matter), you can check on the status of any array using the
# mdadm -D /dev/md0
You get a nice display that tells you how much has been completed, as well as information about the array. As easy as this all sounds, I'd rather have a hardwar RAID controller that rebuilds without any commands needed. With that in mind, best of luck keeping your data safe.