Metrics that talk on cloud using ganglia

Metrics that “talk” on Cloud using Ganglia
How do we monitor performance on Megam Cloud

We'll Cover
●

Our Experience Using Ganglia

●

How does it work in our Platform

●

Chef cookbooks for metering and setup.

●

Dashboard integration

© 2012-2013 Megam Systems

What is Ganglia
Scalable distributed monitoring system for high-performance
computing systems.
Sends information about your cloud instance.
Can be used as a live cloud monitor.
Can be extended using python plugins


What have we accomplished ?
Oh Yeah - <flip to next page>


Dash #2
(AngularJs Client)

Metrics API
gmon

gmon

gmetad

gmetad

gmetad
gmon

gmon


Would you like to setup one ?
Yes you can
(or) http://www.megam.co


Few facts on gmetad/gmond
gmetad can run standalone or along with gmond.
gmetad can be configured to collect metrics of gmond
servers of same cluster or different cluster.
gmetad stores data at
➔

/var/lib/ganglia/rrds/CLUSTER_NAME/GMOND_SERVER_NAME
GMOND_SERVER_NAME can be changed in gmond.conf


gmetad is the metrics collector
gmond is the metrics sender
Ok. Got it.


What is our setup
#1 gmetad : monitor1.megam.co
#2 gmetad : montior2.megam.co
Several gmonds(Cloud Apps) pumping data to gmetad


What are Cloud Apps
Any app
for lang := range ProgLanguages {
Java
Scala
Go
….
meteor

}
&&

DB, Queue

Do you need Graphite
No
Why ?
Needs rrds formatted metric files
It copies rrds files from gmetad
Twice storage
Is this the only soln ? Eager to hear feedback.

Let us setup gmetad 3.3.8-1
Ubuntu(raring) : package is gmetad
Ubuntu(saucy) has 3.6.0

sudo apt-get install gmetad
sudo apt-get install ganglia-webfrontend (*optional)

We used Opscode cookbook to setup => Link


Configure gmetad
nano /etc/ganglia/gmetad.conf
data_source “megcluster” <gmond1>.megam.co:8649 <gmond2>.megam.co:8649

➔

The above says “megcluster” collects metrics from <gmond1>.megam.co and
<gmond2>.megam.co
➔

Which is like monitoring a Java App in <gmond1>.megam.co

(or)
➔

Your favorite App in <gmond2>.megam.co


gmetad - start/stop.
Start :
sudo gmetad

Stop : good old kill
ps -ef | grep gmetad

sudo kill -9 <pid>


Cool gmetad - monitor1.megam.co is running


gmond
Install ganglia-monitor-python package will be installed in a server which is
to be monitored.
➔

Package has methods to collect basic metrics(cpu...) using python scripts
in /usr/lib/ganglia/
➔

Extended by enabling additional python scripts at
/usr/lib/ganglia/python_modules.
➔

For an exhaustive list : https://github.com/ganglia/gmond_python_modules

➔


Let us setup gmond 3.3.8-1
Ubuntu(raring) : package is ganglia-monitor-python
Ubuntu(saucy) has 3.6.0

sudo apt-get install ganglia-monitor-python

We used Opscode cookbook to setup => Link


Configure gmond
nano /etc/ganglia/gmond.conf
globals {
override_hostname = <gmond1>.megam.co
override_ip = 127.0.0.1
}
udp_send_channel“megcluster” collects metrics from <gmond1>.megam.co and
➔
The above says {
host = monitor1.megam.co
<gmond2>.megam.co
port = 8649
ttl ➔ 1
= Which is like monitoring a Java App in <gmond1>.megam.co
}
(or)
cluster {
nameYour favorite App in <gmond2>.megam.co
= "megcluster"
➔
owner = "unspecified"
}

What did we configure ?
➔

In the globals we say our monitoring Java App's host name < gmond1>.megam.co

➔

We provide the UDP channel of the gmetad (monitor1.megam.co)

➔

We need to specify the gmetad cluster (megcluster)

➔

cluster attribute groups all gmond to a gmetd <CLUSTER> in our case
megcluster.


gmond - start/stop.
Start :
sudo gmond

Stop : good old kill
ps -ef | grep gmond

sudo kill -9 <pid>


Cool gmond - <gmond1>.megam.co is running
&
pumping to monitor1.megam.co


We customized chef - cookbooks


How do we use the chef - cookbooks


Tweak cookbook for gmetad

https://github.com/indykish/chef-repo/tree/master/cookbo
➔

Attributes
default[:ganglia][:cluster_name] = "megcluster"
default[:ganglia][:unicast] = true
default[:ganglia][:hostname] = “monitoring1.megam.co”


Chef Run : gmetad
Run chef :
runlist 'recipe[megam_ganglia::gmetad]'


Tweak recipes for gmond
For Any App
➔

Default : installs and configures ganglia-monitor-python. It collects the basic
meterings like cpu_usage, memory_usage etc.

➔

Nginx : collects nginx status details.

➔

Rabbit : collects rabbbitmq metrics.

➔

Redis : collects redis metrics.

➔

Riak

: collects riak metrics.


Chef Run : gmond
To monitor an app
include_recipe “megam_ganglia”
Nginx frontended apps
include_recipe “megam_ganglia::nginx”
Rabbitmq apps
include_recipe “megam_ganglia::rabbit”
Riak apps
include_recipe “megam_ganglia::riak”
Redis apps
include_recipe “megam_ganglia::redis”

Configure gmond
nano /etc/ganglia/gmond.conf
globals {
daemonize = yes
setuid = yes
user = nobody
debug_level = 0
max_udp_msg_len = 1472
mute = no
deaf = no
host_dmax = 86400 /*secs */
cleanup_threshold = 300 /*secs */
gexec = no
send_metadata_interval = 30
override_hostname = "<gmond1>.megam.co"
override_ip = 127.0.0.1
}
cluster {
name = "megcluster"
owner = "unspecified"
latlong = "unspecified"
url = "unspecified"
}
/* The host section describes attributes of the host, like the location */
host {
location = "unspecified"
}
/* Feel free to specify as many udp_send_channels as you like. Gmond
used to only support having a single channel */
udp_send_channel {
host = monitoring1.megam.co
port = 8649
ttl = 1
}


Verifying gmond data

Open up your browser http://gmond1.megam.co:8649
➔

It will list the metrics of the gmond instance.

➔

Not recommended for prod.


Sample gmond data (for redis server)


Checking gmeta data

Data files in the below dir increases
–

Get metrics data at /var/lib/ganglia/rrds/megamcluster/gmond1.megam.co


Sample gmetad data (for thomas.work.local)


Dash Integration in rails

Built on

–
–

We'll cover it detail in a separate slideshare.
If you are hungry “Code is the design” :)
For questions on this area:rajthilak@megam.co.in


References
Ganglia Wiki
megam chef-repo


Our Organization(Megam Systems)
Beta Launch of Megam Cloud (Polygot PaaS)
Our PaaS design => Link
Register http://www.megam.co for an invite
Twitter : @indykish

Screencast illustrating the Cloud API
Servers working live


Thank you

for watching

Metrics that talk on cloud using ganglia

Recommended

More Related Content

Similar to Metrics that talk on cloud using ganglia (20)

Recently uploaded (20)

Metrics that talk on cloud using ganglia