From Funtoo
Revision as of 16:51, 3 November 2010 by Brantgurga (Talk | contribs)

(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to: navigation, search

Utilization monitor
drobbins>	the idea is this - log in to a remote system, grab utilization/capacity data, and exit.
	<brantgurga>	which is what?
	<drobbins>	things like disk, disk io, system load, memory use, network traffic, swap reads/writes
	<drobbins>	this will be for an eventual funtoo network operations center
	<slashbeast>	dstat your friend
	<slashbeast>	it will provide you all this data + more
	<drobbins>	then the data should be logged to a sqlite db
	<slashbeast>	easy to parse and do graphs
	<drobbins>	oh yeah?
	<drobbins>	cool
	<slashbeast>	ya rly
	<drobbins>	ideally no apps need to be instaleld.
	<drobbins>	but for our use, we can require a remote app
	<slashbeast>	impossible.
	<brantgurga>	I think SSH needs to be there, haha
	<drobbins>	yep
	<drobbins>	well, no special apps is what I mean.
	<slashbeast>	-------cpu0-usage--------------cpu1-usage--------------cpu2-usage--------------cpu3-usage------ --dsk/sda-- --net/eth1- ---paging-- ---system--
	<slashbeast>	usr sys idl wai hiq siq:usr sys idl wai hiq siq:usr sys idl wai hiq siq:usr sys idl wai hiq siq| read writ| recv send| in out | int csw 7 2 89 1 0 0: 7 2 89 1 0 0: 7 2 89 2 0 0: 7 2 89 2 0 0| 195k 189k| 0 0 | 226B 1485B| 848 3057
	<slashbeast>	0 1 98 0 0 1: 3 96 0 0 0: 8 3 89 0 0 0: 1 1 98 0 0 0| 0 0 | 12k 740 3B| 0 0 | 652 3096
	<drobbins>	ideally, as much as possible, it should interrogate the system.
	<slashbeast>	a little spammy, sorry
	<slashbeast>	example
	<slashbeast>	it can show diskwrite per partition/device etc
	<slashbeast>	cpu, network, load, iowait
	<slashbeast>	why you need something like it?
	<drobbins>	monitor our infrastructure.
	<slashbeast>	there is a collectd, munin, cacti, nagios
	<drobbins>	yeah.
	<slashbeast>	I hate them all to be honest.
	<drobbins>	basically to see load, etc.
	<drobbins>	maybe not necessary now but as we get more servers it will be necessary.
	<drobbins>	I also have some cool stuff to hook the data into.
	<drobbins>	really I only need samples every 5 min.
	<drobbins>	maybe that's too low?
	<drobbins>	10 sec?
	<slashbeast>	looks ok.
	<brantgurga>	haha, you can create a mediawiki bot to update [[Internal:Load]] with the data
	<drobbins>	two ways to slice it - monitoring (frequent samples), and for capacity (12h slice)
	<drobbins>	here we are doing 12h samples to monitor the utilization of equipment.
	<drobbins>	so for every 12h, how much IO did it do? what was avg/min/max CPU? etc.
	<drobbins>	this is also critical for hosting.
	<slashbeast>	check collectd
	<slashbeast>	you may really like it.
	<drobbins>	ok