2007-06-21
Jason Schoonover
http://unnoc.org

Unnoc Disk Monitoring

Unnoc will graph and log the following (if it is available):

	- Disk Usage
	- Disk I/O Counters

Disk I/O counters are currently only available on Net-SNMP hosts
('server').

Disks are found in two ways:

	1. UCD-DISK MIB
	2. HOST-RESOURCES MIB

You can quickly see if your server supports either method, or you can
see which one would be more appropriate by running two of the following
commands:

	print_disks.pl
	print_disk_io_avail.pl

Both of these scripts are located in the scripts/ directory in the Unnoc
distribution tarball

1. UCD-DISK MIB

This is the disk table that is provided by Net-SNMP when you use the
'disk' directive in snmpd.conf.  It usually comes along with a minimum
value stating that if the disk free value falls below this particular
minimum, then it will raise an alert, and SNMP will set an 'alert' bit
which Unnoc will pick up on and send out appropriate emails.

This is the default method for all 'server' types.

One particular limitation of this method is that usually SNMP will not
report correctly on disks that are greater than 2.2TB.

All minimum disk space requirements are configured inside the snmpd.conf
file.

Disk counters are picked up automatically, if available.  However, if
the disk counters are not picked up correctly, you can manually specify
with the disk{} directive. 

For example, let's say that '/' is an RAID array /dev/md0, however SNMP
does not report anything about /dev/md0.  Then you can specify the
device with the 'device' option inside disk{}.

	host {
		hostname = server1
		type = server
		community = public

		disk {
			partition = /
			device = /dev/md0
		}
	}

Device can also be a comma separated value.

For instance, if you have a partition '/srv' that is one large LVM
partition, and the LVM consists of /dev/sda1, /dev/sda2, /dev/sda3 and
/dev/sda4.  Then you can combine them with the 'device' option and
specify a device_display as well.

	host {
		hostname = server1
		type = server
		community = public

		disk {
			partition = /srv
			device = /dev/sda1, /dev/sda2, /dev/sda3, /dev/sda4
			device_display = /dev/array0/vol0
		}
	}

If you do not specify disk{}, the I/O counters will attempt to be
automatically read via SNMP.


2. HOST-RESOURCE

If you have disks larger than 2.2TB, use this method.

The host resource table is a different MIB.  It allows for more
variables to be read, for devices such as floppies, memory, cpu, etc.
It does not supply disk I/O counters though.

If you are using the HOST-RESOURCE method, then you need to tell Unnoc
that with the 'disktype' directive:

	host {
		...
		disktype = hr
	}

It will automatically read ALL mount points in, sometimes including NFS
partitions.  You can ignore partitions with the 'ignore' directive.  If
you want to include disk counters, you MUST manually specify the device.
If you do not, it will not pick them up.

For alerts, you also MUST manually specify high/warn usage.

	high_usage - a threshold of DISK USAGE that, if met, will raise
		     an alert

	warn_usage - a threshold of DISK USAGE that, if met, will raise
		     a warning

If we have a server with a large array, we need to specify all disk
usage using the disktype = hr.

For example:

	host {
		hostname = server2
		type = server
		community = public
		disktype = hr

		disk {  
			partition = /proc/bus/usb, /mnt/nfs
			ignore = 1
		}

		disk {
			partition = /
			high_usage = 90%
			warn_usage = 80%
			device = /dev/sdb3
		}

		disk {
			partition = /boot
			high_usage = 90%
			warn_usage = 80%
			device = /dev/sdb1
		}

		disk {
			partition = /srv
			high_usage = 95%
			warn_usage = 90%
			device = /dev/sda1, /dev/sda2, /dev/sda3, /dev/sda4
			device_display = /dev/array0/vol0
		}
	}


Remember that device can be a comma separated value, and if you
specify multiple devices, then you can use the device_display
directive.


vim:tw=72:wm=1