My Production EM12c Upgrade From R2 (12.1.0.2) to R3 (12.1.0.3)

This post covers my production upgrade from EM12c R2 to EM12c R3 on Linux x86-64 (SLES11 SP2).  To stress test the upgrade and keep it interesting, this system also has BI Publisher integrated into OEM, and also has the plugin from NetApp (version 12.1.0.1.0), the VMware plugin from BlueMedora (version 12.1.0.5.0), and the MySQL plugin from Pythian (version 12.1.0.1.1).  The repository database is running on version 11.2.0.3.

I’m feeling lucky today.  So this is just going straight into production.  Famous last words…

Preparation

  1. Go to edelivery and search for the Oracle Enterprise Manager product pack and platform Linux x86-64
  2. Follow the link titled “Oracle Enterprise Manager Cloud Control 12c Release 3 (12.1.0.3) Media Pack for Linux x86-64
  3. Download the three files needed for EM12c R3: V38641-01.zip, V38642-01.zip, and V38643-01.zip
  4. Download patch 13349651 for WebLogic, you will need it during the post-upgrade steps
  5. View the digest and run md5sum against each downloaded file to confirm that the files downloaded correctly
  6. Transfer the EM12c R3 files to a staging area on your Oracle Enterprise Manager server and unzip all three of them
  7. Delete the three downloaded .zip files if you are short on space (but don’t just “rm *.zip” or you’ll remove the necessary WT.zip file at the top level of your staging directory)
  8. Review the upgrade guide
  9. Create backups of your current OMS home, agent home, software library and Oracle inventory
  10. Create a backup of your current repository database that you can restore from if necessary
  11. If you use a dedicated filesystem for your Oracle Management Agents, make sure that dedicated filesystem has enough free space.  I used a small, 2GB filesystem, and this was barely large enough, except for the one sandbox server where it was too small and I was not able to complete the agent upgrade without adding space
  12. Stop the BI Publisher WebLogic Server (if you have it installed) via the WebLogic Admin Console
  13. Make sure your repository database does not have snapshots created on any tables by running “select master, log_table from all_mview_logs where log_owner = ‘SYSMAN'” — if any snapshots are found, follow the instructions in the upgrade guide to drop them
  14. If your repository database is version 11.1.0.7 or 11.2.0.1, follow the steps in the upgrade guide to apply the prerequisite patches needed to proceed
  15. Copy the emkey from the existing OMS to the existing management repository by running “$OMS_HOME/bin/emctl config emkey -copy_to_repos” and enter the SYSMAN password when prompted
  16. Confirm the emkey was copied by running “$OMS_HOME/bin/emctl status emkey” and enter the SYSMAN password when prompted
  17. Stop every OMS in your environment by running “$OMS_HOME/bin/emctl stop oms -all”
  18. Stop the management agent monitoring the management services and repository target by running “$AGENT_HOME/bin/emctl stop agent”

Running The Upgrade

Due to some issues I had with the upgrade from EM12c R1 to R2, I highly recommend that you do NOT use Cygwin to ssh to your OMS host and display the installer over ssh using Cygwin’s X server.  Use VNC instead.  I’m not going to try Cygwin this time through.

First, I start vncserver on the OMS host.  Then I connect to it to using TightVNC from my desktop machine.

  1. Navigate to the staging directory where you unzipped the EM12c R3 distribution files and run “./runInstaller” as your oracle software owner
  2. As an SAP customer, we are not allowed to use OCM, so I skipped the steps involving entering my My Oracle Support credentialsStep 1
  3. I also skipped the search for updates as this is a new enough release at the moment there should not be any necessaryStep 2
  4. The prerequisite checks run.  I received a warning about libstdc++43-4.3 not being found, but libstdc++43-devel-4.3.3-x86_64 fulfills this need, so I click ignore and then Next to continueStep 3
  5. Only the one system upgrade is supported when upgrading from EM12cR2, so I selected “Upgrade an existing system”, “One system upgrade”, and my existing middleware home.  Click NextStep 4
  6. This out of place upgrade will go into a new middleware home.  I am using /oracle/oem/Middleware12cR3.  Click Next and the installer will confirm that you have enough free space availableStep 5
  7. Here I had to pause and request more space from my storage admin, as the installer wants at least 14.0GB of free space.  Once that was done, I proceeded
  8. Enter the SYS and SYSMAN passwords and check the box to confirm that you have backed up the repository (you should have your OMS, agent, etc all backed up as well), then click NextStep 6
  9. The installer will check various parameters on your repository database and offer the chance to fix them if any need to be changed.  I accept the fixes and click YesStep 6(b)
  10. The installer checks some additional settings and notes that they should be reviewed after the installation or fixed now.  I explicitly granted execute on DBMS_RANDOM to DBSNMP and then clicked OKStep 6(c)
  11. The installer lists the plugin versions that will change and the plugins that will migrate. Confirm this all looks right and then click NextStep 7
  12. The installer lists additional plugins you can choose to deploy at install time.  I do not use any of these so I left them all unchecked and clicked NextStep 8
  13. The installer requests the password for your WebLogic adminserver and confirmation of the hostname, port and username.  Provide the password and click Next.  You may be able to change the OMS instance directory here but I do not suggest doing soStep 9
  14. You now have a chance to review your settings, then click Install to proceed with the upgradeStep 10
  15. Installation proceedsStep 11
  16. You are then prompted to run the allroot.sh file.  Login to the server and execute it as root or via sudoallroot.sh
  17. Once the install/upgrade is complete, the installer will display an installation summary.  Review it, save the URLs it gives you for the OMS and adminserver, then click CloseUpgrade Summary Report

Overall, the upgrade installation steps took 1 hour and 15 minutes in my environment.  This was on a physical server with 126GB RAM, 16 dual core processors and 200 managed targets.  This does not include the post-upgrade steps shown below.

Post Upgrade Steps

Now that the upgrade is complete, return to the upgrade guide to complete post upgrade steps.  Your environment may differ from mine, but these are the steps I had to follow.

  1. Start your central agent by running “$AGENT_HOME/bin/emctl start agent”.  At this point the load on my system went up very high and began responding very slowly.  I walked away for 10 minutes to let things settle down
  2. Open your web browser and go to the URL provided at the end of the installation and login as SYSMAN.  When I first tried to do so using Firefox I received an error indicating an invalid certificate.  I had to delete the old certificates and authorities from my previous installation and restart Firefox before it would allow me in.  MSIE worked fine though
  3. Update the central agent (the management agent installed on the OMS host) by clicking on the Setup menu, then Manage Cloud Control, then Upgrade Agents.  Click the Add button and select your central agent.  I choose “Override preferred credentials” since I have not configured sudo.  Click Submit to continue, then OK when warned that you may have to run root.sh manuallyUpgrade Central Agent
  4. My first upgrade attempt on the central agent failed due to a prerequisite check for package libstdc++-43.  The easy thing to do here is expand the Additional Inputs region and provide “-ignorePrereqs” as an additional parameter, but I chose to complete this agent upgrade using emcli and describe that process in the next two steps
  5. First run “$OMS_HOME/bin/emcli login -username=sysman” and enter the SYSMAN password when prompted.  Then run “$OMS_HOME/bin/emcli sync”
  6. Upgrade the agent by running “$OMS_HOME/bin/emcli upgrade_agents -additional_parameters=”-ignorePrereqs” -agents=”hostname.domain.com:3872″
  7. Wait while the upgrade proceeds.  You can view upgrade progress by running “$OMS_HOME/bin/emcli get_agent_upgrade_status”, or in the GUI by clicking on “Agent Upgrade Results” in the “Upgrade Agents” pageAgent Upgrade Progress
  8. Click Done once the central agent upgrade completes.  Go to the new agent home and run root.sh as root or via sudo
  9. Then repeat this process for the rest of your agents.  Try to install them first WITHOUT using the “-ignorePrereqs” flag, because if there are missing prerequisites you need to identify the issue and find out if it is something that it is appropriate to ignore, as the libstdc++ version was in my case.  Execute root.sh for each agent afterwards, unless you have sudo configured in which case it will happen automatically
  10. Two of my agents that run on different platforms could not be upgraded right away.  The new versions of the agent software needed to be downloaded from Self Update.  I am skipping them for now
  11. Next, apply patch 13349651 to WebLogic, following the instructions in the README file.  I attempted to do so, but the patch was already installed so I skipped this step
  12. There are a few other optional, post-upgrade steps like deleting obsolete targets.  These are documented in the upgrade guide and I will not note them here
  13. As a final step, make sure you update your Oracle user’s environment variables to reflect the new middleware home, OMS home, agent home, and so on

Conclusion

At this point my EM12cR3 production upgrade is complete!  Everything I have checked so far appears fully functional.  The only problem I had was a small filesystem for the management agent on my sandbox server causing the agent upgrade to run out of space, forcing manual intervention to resolve.  Don’t be stingy with space like I am and you should be fine.

I haven’t taken any time to investigate the new features yet, but I will be now.

(EDITED TO ADD: I forgot to mention the steps to get BI Publisher working again.  Please refer to the EM12cR3 Advanced Installation and Configuration Guide, chapter 15.  Essentially you will need to perform a software-only installation into the new middleware home, then execute the configureBIP script with an -upgrade flag to complete the BIP setup.)

Advertisements

12 thoughts on “My Production EM12c Upgrade From R2 (12.1.0.2) to R3 (12.1.0.3)

  1. Pingback: My Production EM12c Upgrade From R2 (12.1.0.2) ...

  2. Pingback: How to migrate EM12c R3 OMS and repository to a new host | Pardy DBA

    1. Brian Pardy Post author

      Hi,

      Thanks for commenting.

      Eventually, yes you can delete the old OMS folder, but you should follow the procedure in the documentation to use the OUI tool to deinstall the old components.

      The process is documented here:

      http://docs.oracle.com/cd/E24628_01/upgrade.121/e22625/appdx_delete_old_oms_home.htm#CHDHIHCC

      By following this procedure you can make sure that the components from the previous version are all properly uninstalled and unregistered from the Oracle inventory. If you just delete the folder without uninstalling, you may have issues later with discovery on that server, or applying patches, or other problems. I have always followed the deinstall process from the upgrade guide.

      Hope this helps!

      Reply
  3. Pingback: Slow DB Home Page EM12c | Mehmet Eser's Oracle Blog

  4. Damir Vadas

    >do NOT use Cygwin to ssh to your OMS host
    In any normal server, you CANNOT install VNC.
    And Cygwin is officially supported by Oracle (whole EBS on Win is running inside it!)….so I think you have misconfiguration ssh…or I understand something wrong.

    Reply
    1. Brian Pardy Post author

      Hi Damir,

      I think you understand everything correctly and I agree with all of your statements. I use Cygwin to connect for all of my installations and for nearly all of my maintenance work.

      However, I ran into a bug during the upgrade from EM12cR1 to EM12cR2, and the bug was due to Cygwin’s X windows timeout, and solved by using VNC to connect when running the upgrade.

      See https://community.oracle.com/thread/2441070?start=0&tstart=0 for more information, also bug number 14623310.

      I believe the bug is now patched but for superstitious reasons I will use VNC for every OEM upgrade going forward.

      Reply
  5. Mahesh Padhmanabhan

    Pardy, if i am right, you did a one system upgrade. we are also planning to upgrade from 12.1.0.2 to 12.1.0.3 . ours is a two node RAC with similar setup in dataguard. we have around 7000 targets.
    database size is 120G
    will the upgrade time increase with more number of targets. yours took 1 hr 15 min for 200 targets.
    how much time do you think our upgrade will take.

    Reply
    1. Brian Pardy Post author

      Hi Mahesh,

      You’re correct, this was a one system upgrade.

      That’s a great question. I’m reaching out to some other people to find out if they know. I would assume that the upgrade will take longer with more targets, since some tables have their definitions changed and some data has to be converted, but I expect your upgrade time will mostly depend on the speed of your storage and CPUs compared to mine, rather than the target count.

      Reply
  6. khalid

    Installing OEM on Linux x86-64 can be more reliable and robust than on AIX platform , you always face some issues or bugs. we faced an issue on a fresh install of OEM13 on aix so we want back to 12cr5.
    is 13c solid version so far ?

    Reply
    1. Brian Pardy Post author

      13c has been very solid for me so far, but I only use it for database management. I do not have it managing any middleware or server deployments or using very many of the advanced features outside of database management. I have only run into one small issue, which happened while applying patches to agents through OEM. Patch application fails if I check the box to “Upgrade OPatch version” when applying a patch plan to an agent, because EM13c uses a version of OPatch that is not yet available on MOS. All I had to do was leave that box unchecked during deployment and everything has worked successfully.

      Reply

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s