StarCluster - Mailing List Archive

Re: Ganglia Plugin ?

From: Sergio Mafra <no email>
Date: Mon, 25 May 2015 11:14:07 -0300

Hi SC_Folks,

I´ve made a little research and found interesting things that helped to
build a plugin of configuring Ganglia to run with StarCluster.
Here comes what I did in case if you´re interested:

1. First, I used the Shared StarCluster AMI Ubuntu 14.04. It can be found
on Public AMIs. (us-east-1 -> ami-38b99850).
2. This tutorial was great:
3. I´ve installed these apps in base AMI using: sudo apt-get install -y
ganglia-monitor rrdtool gmetad ganglia-webfrontend
4. Ganglia needs to be configured using two config files in /etc/config:
gmetad.conf and gmond.conf
5. The gmetad.conf needs to have the name of the cluster you intend to
monitor. I used a fixed name (StarCluster) but you can chance in the future
with the name tag of the cluster you´re provisioning, So this is what I´ve
data_souce "StarCluster" 60 localhost
6. The gmond.conf needs to changed in 3 sections:
 a) Cluster section: Note that I´ve change the unspecified to StarCluster
(same name in the gmetad.conf)
 cluster {
   name = "StarCluster"
   owner = "unspecified"
   latlong = "unspecified"
   url = "unspecified"

b) udp_send_channel section: You have to comment the mcast_join line and
add the ip of the cluster master. The plugin will handle this. So I add the
line "host = master_ip". This will be the same if the node is master or
udp_send_channel {
# mcast_join =
  host = master_ip
  port = 8649
  ttl = 1

c) udp_recv_channel: You will have to have two kinds of configuration
regarding the master and the slaves. In master, the section will be the
following: The mcast_join line commented and bind line too.

> Master node:
udp_recv_channel {
# mcast_join =
  port = 8649
# bind =

> Slave node: All section must be commented:
#udp_recv_channel {
# mcast_join =
# port = 8649
# bind =

So I decided to create two gmond.conf in base AMI. The first one will be
used to all Slaves and the second will be used to Master.
This will be changed in Plugin time.

I´ve prepared the gmond.conf with a) , b) and c) Slave node and
gmond_master.conf with a), b) and c) Master node.

With the ganglia plugin:

1. Gets the private ip of the Master node (master_ip)
2. Checks if the node is master.
3. If true, replaces the gmond.conf file from gmond_server.conf, replaces
the master_ip in this file and restart the services needed.
4. If false, replaces the master_ip in gmond.conf and restart the
ganglia-monitor service.

from starcluster.clustersetup import ClusterSetup
from starcluster.logger import log

class ganglia(ClusterSetup):
     def __init__(self):
          log.debug('Installing Ganglia...')
     def run(self, nodes, master, user, user_shell, volumes):
  "Installing Ganglia in all nodes")
            master_ip = None
            for node in nodes:
                if node.is_master():
                   master_ip = node.private_ip_address
                   master.ssh.execute('mv /etc/ganglia/gmond_server.conf
                   master.ssh.execute("sed s/master_ip/%s/
/etc/ganglia/gmond.conf > ar&&mv ar /etc/ganglia/gmond.conf" % master_ip)
                   master.ssh.execute('service ganglia-monitor restart &&
sudo service gmetad restart && sudo service apache2 restart')
                   node.ssh.execute("sed s/master_ip/%s/
/etc/ganglia/gmond.conf > ar&&mv ar /etc/ganglia/gmond.conf" % master_ip)
                   node.ssh.execute('service ganglia-monitor restart')

That´s it.. You now can check on http://<your_master_ip_address>/ganglia
and monitor all cluster´s stuff.

Hope this helps you,

Sérgio Mafra

On Fri, May 22, 2015 at 8:47 AM, Sergio Mafra <>

> Hi SC_Folks,
> Is there any plugin around developed to install Ganglia on StarCluster?
> All best,
> Sergio
Received on Mon May 25 2015 - 10:14:09 EDT
This archive was generated by hypermail 2.3.0.


Sort all by: