I'm with you that the documentation is (at least) wrong.
DISCLAIMER: I do not own a RHEL system. I have extensive experience of Debian raid, from different Debian versions, but definitely never put my hands on a RHEL system. Please test this on a spare system (VM) before acting on a production machine.
This is what will happen:
Remove the disk from the raid array.
mdadm --manage /dev/md0 -r /dev/sdc3
The sdc3 partition will be expelled from the md0 array.
- Remove the disk from the system.
The entire sdc disk, amoung with its sdc3 partition, will not be anymore in the system, I assume "they" want us to simply "sit the failed sdc disk on the desk" for now.
- Using fdisk, replace the removed disk and re-format the replacement disk.
You connect a new sdc disk, then you re-partition it exactly like the failed one. At this point, a "sdc3 partition will exist again in the system
Add the new disk back to the RAID array.
mdadm --manage /dev/md0 -a /dev/sdc3
The sdc3 partition will be added to the array, and a resync will start automatically, and immediately.
This is simply what it's needed to replace a failed disk. You simply have to wait for the resync to finish, by watching /proc/mdstat, and nothing more.
But the documentation goes on.
To restore the disk, perform a "software fail" the previous spare slice:
mdadm --manage --set-faulty /dev/md0 /dev/sdc3
This will set the sdc3 partition as faulty, simulating a hardware fail. You will go back to before step 1, when you had the sdc3 partition failed.
The system will now attempt to rebuild the array on the replaced disk. Use the following command to monitor status:
watch -n 1 cat /proc/mdstat
Actually, no. The system will wait for your intervention, and you should now follow the steps 1-4 again.
However, this command is the right one for monitoring the resync progress, with a 1-second updates. Press CTRL+C to exit from the watch process.
When the array is finished rebuilding, remove and then re-add the software-failed disk back to the array.
mdadm --manage /dev/md0 -r /dev/sdc3
mdadm --manage /dev/md0 -a /dev/sdc3
Actually, these two commands will make the rebuild (resync) process start. They are exactly step 1 and 4 (steps 2 and 3 are not needed now, since the disk has not failed because of a real problem, but because of a simulated error).
You now have to wait for the rebuild process to finish, monitoring /proc/mdstat.
Check the array.
mdadm --detail /dev/md0
This will give you more info about the md0 array, but nothing more about the resync process which you can already get from /proc/mdstat. Not needed at all.
/proc/mdstat
that the array is rebuilding or rebuilt, then you do not need to perform them.