Monday, 10 March 2014

EM12c and Exalogic 2.0.6

Introduction

In an earlier posting I wrote about the process used to integrate a virtualised Exalogic rack with EM12c.  Since that article was written both EM12c and Exalogic have had an upgrade.  This posting is a short one to highlight the changes done/work needed to get the following combination of products working together:-
  • Exalogic - 2.0.6.n.n
  • EM12c - 12.1.0.3.0
In EM12c the plugins now used are:-
  • Oracle Virtualization - 12.1.0.5.0
  • Oracle Engineered System Healthchecks - 12.1.0.4.0
  • Sun ZFS Storage Appliances - 12.1.0.4.0

Agent Deploy

The steps to deploy the agent to EMOC and OVMM is the same as previously mentioned.  The changes from the previous instructions are:-
  • Now EMOC and OVMM are both on the same vServer so there is only one agent that needs to be deployed.
  • No need to create the /var/exalogic/info/em-context.info file.  However you do need to ensure the /var/log/cellos/SerialNumbers file is created and populated.  (See deployment notes)
  • The sudo permissions can be simplified down to the oracle user permissions of:-
    oracle ALL=(root) /usr/bin/id, /opt/oracle/em12c/agent/*/agentdeployroot.sh
Note 1 - I have done this agent deployment a number of times now and because the labs I am working in often do not have DNS fully setup I end up using the local hosts files for name resolutions.  It is critical that both the OMS server and the agent deployment target server have the target hostname fully qualified in their hosts files.  Otherwise the latter steps of deployment when EM12c attempts to "Secure Agent" will fail to startup the agent.

Note 2 - The same process applies to deploying an agent to a guest vServer.  However, be warned I did this on a vServer using the base 2.0.6.0.0 guest based template that had a couple of other small applications on it.  The agent deployment uses a reasonable amount of disk space (~1Gb) and during the deployment this can fail.  The log on the OMS server was reporting an error "pty required false  with no inputs".  It turns out that this was because the first step in the "Remote Prerequisite Check Details" performs an unzip of the installation media which was running out of disk space and hanging.  Killing the unzip process caused the step to fail and indicate that the likely cause was disk space.  To avoid this in the first place ensure there is adequate disk space on the vServer.

Note 3 - In order to ensure that a vServer is discovered correctly as part of the Exalogic rack it is necessary to ensure that the file /var/log/cellos/SerialNumbers is generated from the dmidecode command output.  The script shown below can be used to generate this. Simply cut and paste this into a file called generateSerialNo.sh, make it executable and run it on the vServer.


   serialCode=`dmidecode |grep Serial|grep -v Not|cut -d ":" -f2|cut -d " " -f2`
    if [ -f /var/log/cellos/SerialNumbers ]; then
        echo "File /var/log/cellos/SerialNumbers already exists."
    else
        mkdir -p /var/log/cellos
        echo "====START SERIAL NUMBERS====" > /var/log/cellos/SerialNumbers
        echo "==Motherboard, from dmidecode==" >> /var/log/cellos/SerialNumbers
        echo "--System serial--" >> /var/log/cellos/SerialNumbers
        echo "$serialCode" >> /var/log/cellos/SerialNumbers
        echo "--Chassis serial--" >> /var/log/cellos/SerialNumbers
        echo "$serialCode" >> /var/log/cellos/SerialNumbers
    fi



Note 4 - The configuration of the agent will fail if it cannot resolve its own hostname to a valid IP address. i.e. Make sure that /etc/hosts has an entry in it that specifies the hostname of the vServer you are adding.  (When creating vServers this does not happen by default as the names put into the /etc/hosts file are <server name>-<Network IP - dash separated>.)

Setup ZFS Appliance target

The setup process for the ZFS appliance is the same as described previously apart from a minor change that is required to the ZFS configuration.  During the process of seting up the appliance as a target it is necessary to run the "Configuring for Enterprise Manager monitoring" workflow.  This creates the user that will be used by the agent to log onto the appliance and gather the stats.  However the user that is created is defined as a kiosk user.  It is necessary to deselect this option from the user created (oracle_agent) because EM12c requires access to some other data from the console.  If this option is not deselected then EM12c discovers the appliance but raises an alert "Cannot monitor target. Incorrect Credentials"

Having corrected the kiosk user then in the EM12c interface select the ZFS target and under Monitoring --> All Metrics select the metrics you are interested in and enable the ones you are interested in.  (By default they are disabled.)

Conclusion

The general process for deployment is essentially unchanged, just a few minor variations on a theme.