I would like to list down the steps on how I do the collectd installation on the client node.
1. Download the relevant RPM. I am choosing the latest collectd version, which is 5.4.1. The default package that came from EPEL repo is on version 4, which is a bit outdated. The reason behind it collectd on version 4 is not comes with the graphite plugin. So, to take the shortcut, without compiling the binary from scratch of tarball, you can ought to be like me, and download/install the ready made RPM available from the internet.
wget http://anfadmin.ucsd.edu/linux/CentOS/6/x86_64/collectd-5.4.1-1.el6.x86_64.rpm
wget http://anfadmin.ucsd.edu/linux/CentOS/6/x86_64/libcollectdclient-5.4.1-1.el6.x86_64.rpm
rpm -ivh libcollectdclient-5.4.1-1.el6.x86_64.rpm collectd-5.4.1-1.el6.x86_64.rpm
vi /etc/collectd.conf
Hostname "centos66"
Host "graphite_hostname"
Port "2003"
Protocol "tcp"
LogSendErrors true
Prefix "collectd."
#Postfix "collectd"
StoreRates true
AlwaysAppendDS false
EscapeCharacter "."
/etc/init.d/collectd start
Please get back to your graphite server, and take a look at the graphite dropdown tree, it should populate your new collectd host. If you want to repeat the same things, on let says, 200 servers? you can always use my pyssh.py script to login to the servers concurrently and make changes and apply the change as you like.
Next I guess, you would like to explore on the presentation layer, like Graphana. With graphana installed, you can show/manipulate the metrics in a narrow scope to fit the need of monitoring needs to you and your team. Or an more intuitive tool like logstash to give you more analytical view on your system log. Or you need a alerting system like graphite-beacon to generate alerts in a different formats.
Congrats!
No comments:
Post a Comment