Matt once wondered (on the dev list) why I don't write documentation. So
after a solid day of SCSI troubleshooting, I thought I'd, you know,
"contribute..."
---
Here are the metrics that are widely supported across different platforms
(or, in a few cases, the ones we *wish* were supported across the different
platforms ... )
boottime
Number of seconds since system last got Das Boot.
cpu_num
Number of CPUs in the system.
cpu_speed
Speed of the CPUs in the system (not guaranteed accurate).
cpu_user
Percentage of CPU cycles spent in user mode.
cpu_system
Percentage of CPU cycles spent in non-user mode.
cpu_nice
Percentage of CPU cycles spent on nice processes.
cpu_idle
Percentage of CPU cycles spent heating your machine room.
cpu_wio
Percentage of CPU cycles spent waiting for I/O (Solaris)
cpu_aidle
Percentage of CPU cycles spent idle since last boot (Linux)
gexec
Is the Ganglia Execution environment running? (Linux)
heartbeat
Google search for phrase: "machine that goes ping"
load_one
Reported system load, averaged over one minute.
load_five
Reported system load, averaged over five minutes.
load_fifteen
Reported system load, averaged over fifteen minutes.
location
Location of the rebel base.
machine_type
The CPU architecture on which Ganglia is running.
mem_buffers
Amount of memory allocated to system buffers (Linux)
mem_cached
Amount of memory allocated to cached data (Linux)
mem_shared
Amount of memory occupied by processes.
mem_total
Total amount of physical memory.
mtu
Smallest Maximum Transmissible Unit value for all
attached, operational interfaces connected.
601 == unsupported (broken on some platforms).
os_name
The name of the operating system on which blah blah blah.
os_release
The version of the operating system / kernel etc.
proc_run
Number of running processes (not on Solaris, IRIX or OSF)
proc_total
Number of total resident processes (not on OSF).
swap_free
Amount of free swap space.
swap_total
Amount of total swap space.
sys_clock
Number of seconds since January 1st, 1970, according to
the local system clock.
Now for the metrics that probably won't get ported widely...
Solaris-specific metrics (should be the equivalent of sar's stats):
bread_sec
Number of reads between buffers and block devices.
bwrite_sec
Number of writes between buffers and block devices.
lread_sec
Number of reads of system buffers.
lwrite_sec
Number of writes to system buffers.
phread_sec
Number of reads using physical devices.
phwrite_sec
Number of writes using physical devices.
rcache
(1-bread/lread), percentage of read cache hits.
wcache
(1-bwrite/lwrite), percentage of write cache hits.
Linux-specific metrics:
bytes_in
Number of bytes read from all non-loopback interfaces.
bytes_out
Number of bytes written to all non-loopback interfaces.
pkts_in
Number of packets read from all non-loopback interfaces.
pkts_out
Number of packets written to all non-loopback interfaces.
disk_total
Total capacity on the fullest local disk partition.
disk_free
Total free space on the fullest local disk partition.
part_max_used
Name of the partition used in disk_total/disk_free.
I don't think this is in the docs. But it could be. *cough* Although
maybe Matt'd wanna change two or three of them.
If this isn't in-depth enough for you, get your hands on the source and
check out $SOURCE_ROOT/gmond/machines - the logic (or lack thereof) behind
each of these metrics for each platform is in $PLATFORM.c ...
Anyway, back to writing my bizarro Ganglia extensions...
