INFN wiki

This page is read only. You can view the source, but not change it. Ask your administrator if you think this is wrong.
====== Monitoring the NFS cluster with Nagios ======

The NFS service provided by the 2 nodes cluster is an "active/passive" one. In this case the normal behaviour is:

  * one node is actually running the nfsd daemon
  * the other nodes are in standby
  * takeover of the service is handled by the cluster daemons 

Therefore we decide to monitor the situation by:

  * checking the **cluster** daemons are running on each node
  * checking the ''nfsclusterserver'' service is running on the cluster
  * if the server is running on my host check detailed status of the **nfs** daemons
  * if I'm a standby node but the cluster is OK and nfs is running somewhere return OK

===== Install needed packages =====

On all the monitored nodes:

<code bash>
# yum -y install nrpe nagios-plugins-perl perl-Nagios-Plugin
</code>

Obtain latest version of the monitoring scripts from [[https://exchange.nagios.org/directory/Plugins/Operating-Systems/Linux/check_nfs4/details|here]]
and [[https://exchange.nagios.org/directory/Plugins/Clustering-and-High-2DAvailability/Check-CRM/details|here]] and copy them on the relevant directory

<code>
# cp check_nfs4.0.2.pl /usr/lib64/nagios/plugins/check_nfs4
# cp check_crm_v0_7 /usr/lib64/nagios/plugins/check_crm
# chmod +rx /usr/lib64/nagios/plugins/check_nfs4
# chmod +rx /usr/lib64/nagios/plugins/check_crm
</code>

Since all nodes on the cluster share the same domain and users we do not use the idmapd daemon. Its absence is therefore not critical:

<code bash>
sed -i 's/^if (!$idmapd_d) { $daelist/# if (!$idmapd_d) { $daelist/' /usr/lib64/nagios/plugins/check_nfs4
</code>

===== Create an helper script =====

To implement the nagios check as designed we use an helper script that checks if the nfs daemon is running on the tested host or not. 
In the former case the result of the check is handled over to the ''check_nfs4'' script:

<code bash check_my_nfs>
#!/bin/bash

monitor="/usr/sbin/crm_mon -1"

# check cluster is healthy
${monitor} -s 1>/dev/null
if [ "$?" != "0" ];
then
   echo "Cluster is not OK!"
   exit 2
else
   #
   # check if there is at least one nfs server active
   #
   ${monitor} | grep nfsclusterserver | grep -i started 1>/dev/null
   if [ "$?" != "0" ];
   then
      echo "NFS server is not running anywhere!"
      exit 2
   else
      hname=$(hostname -s)
      ${monitor} | grep $hname | grep nfsclusterserver 1>/dev/null
      if [ "$?" = "0" ];
      then
         #
         # I am the nfs server: check if I'm healthy
         #
         exec /usr/lib64/nagios/plugins/check_nfs4
      else
         #
         # I am not the nfs server but:
         # - the cluster is ok
         # - the service is running
         #
         echo "NFS is running somewhere..."
         exit 0
      fi
   fi
fi
</code>
===== Setup nrpe on monitored hosts =====

==== nrpe directives ====

On all the hosts composing the cluster create the file ''/etc/nrpe.d/check_nfs4.cfg'' containing the following directives:

<code>
# Allow requests from cld-nagios by adding the cld-nagios IP to the list of allowed hosts
allowed_hosts=127.0.0.1,192.168.60.32

# Define the check_crm command:
command[check_crm]=/usr/lib64/nagios/plugins/check_crm

# Define the check_nfs4 command:
# On CentOS the file '/var/log/messages' is readable only 
# by root so we run this check through 'sudo'
command[check_nfs4]=sudo /usr/lib64/nagios/plugins/check_my_nfs  </code>


==== Allow nrpe to run the checks as root ====

  * Create the file ''/etc/sudoers.d/nrpe'' containing
<code>
Defaults:nrpe !requiretty

nrpe ALL = (root) NOPASSWD: /usr/sbin/crm_mon
nrpe ALL = (root) NOPASSWD: /usr/lib64/nagios/plugins/check_my_nfs
nrpe ALL = (root) NOPASSWD: /usr/lib64/nagios/plugins/check_nfs4 -v
</code>

  * Give the file the correct permissions
<code bash>
chmod 440 /etc/sudoers.d/nrpe </code>


==== Open firewall port 5666 ====

<code bash>
firewall-cmd --add-port=5666/tcp
firewall-cmd --permanent --add-port=5666/tcp
</code>

==== Start and enable the nrpe daemon ====

<code bash>
systemctl start nrpe
systemctl enable nrpe
</code>
===== Define needed commands on cld-nagios =====

  * Make sure nrpe is installed on the nagios server
<code># rpm -qa | grep nrpe
nrpe-2.15-2.el6.x86_64
nagios-plugins-nrpe-2.15-2.el6.x86_64</code>

  * Make sure a command to exec checks using nrpe is defined (check the ''commands.cfg'' file)
<code>
define command{
    command_name            check_nrpe_cedc
    command_line            $USER1$/check_nrpe  -H $HOSTADDRESS$ -t 480 -c $ARG1$
}
</code>

  * Create the new command that execs check_nfs4 on the monitored host
<code>
define command{
        command_name        check_nfs4
        contact_groups      cedc-admins
        command_line        $USER1$/check_nrpe_cedc -H $HOSTADDRESS$ -c check_nfs4
}
</code>

  * Add it to the list of the scheduled checks for every node in the cluster
<code>
define service{
        use                             server-service         ; Name of service template to use
        contact_groups                  cedc-admins
        host_name                       cld-blu-01
        service_description             NFSv4 Status
        check_command                   check_nrpe_cedc!check_nfs4
        }  
</code>

  * Create the new command that execs check_crm on the monitored host
<code>
define command{
        command_name        check_crm
        contact_groups      cedc-admins
        command_line        $USER1$/check_nrpe_cedc -H $HOSTADDRESS$ -c check_crm
}
</code>

  * Add it to the list of the scheduled checks for every node in the cluster
<code>
define service{
        use                             server-service         ; Name of service template to use
        contact_groups                  cedc-admins
        host_name                       cld-blu-01
        service_description             CFS Cluster Status
        check_command                   check_nrpe_cedc!check_crm
        }  
</code>
===== Reload nagios =====
<code>/etc/init.d/nagios reload</code>