Fyod Administration

FYOD is a persistant demon that runs one copy on each of the I/O nodes.

In addition to starting FYOD on the I/O nodes, FYOD requires four things from the system administrator.

There must be line in the host file that defines FYOD. As seen by the compute node the file in question is
```
	/cplant/cplant-host
```
The service is "fyod", and ALL FYOD nodes must be listed, in the form c-n.SU-m, where n and m are correctly choosen. (If an I/O node is temporarily removed from service this file does NOT have to be editted. There will be occasional small delays seen by the app, if extra nodes are in the list, but it should not prevent operations.) However, the first node in the list must be operational, for FYOD files services to be available. This is the only address, YOD passes to the app.
Presently hard coded in FYOD is the assumption, that the Raids are mounted as /etc/var/sfyod0 and /etc/var/sfyod1. (If one or both is not mounted, it will be gracefully omitted, but it is, of course, inaccessible.)
Each Raid must have a file at that lowest level named
```
	.raid_master,  for example,
	/etc/var/sfyod0/.raid_master
```
The contents of that file is two numbers. The first is a "unit number" which is of the form xxi, where i is either 0 or 1. These numbers must be unique across the raids! (There are implementation deficiencies at the moment, but the intent is that a raid could be mounted to any IO processor box and it would retain its unit number. That is, the compute nodes and the application program do not have, a priori, knowledge of where a particular raid unit is located.) The second number is called a validation check, but is not checked at this time. Such a file (for unit 51) can be created by
```
	echo 51 773355 > /etc/var/sfyod1/.raid_master
		DO NOT make it "051" as that gets read as octal
                   and hence comes out as 41 which is not what is
                   expected!
```
At least for now, the last digit, the 0 or 1, should match between the unit number and sfyod0 or sfyod1.
Note that this is written ONCE, NOT each time we boot. This is, in effect, the identification of the raid. (Probably write permission, even for root, should be removed from this file.)
The directory tmp needs to be created on each raid also and given 777 permissions.
```
	mkdir /etc/var/sfyod0/tmp
	chmod 777 /etc/var/sfyod0/tmp
			(sfyod0 or sfyod1 as the case my be.)
```
Likewise, this is done once.
In the present 2.0.35, release 0.3, it would appear that all that can be supported is one unit per IO node. Hence, there is only /etc/var/sfyod0 to consider. The unit numbers should then be simple, like 10, 20, 30, 40, 50, 60.