Monday Oct 19, 2015

OOW 2015: CON8833 Best Practices from Oracle Development for Patching and Upgrading Oracle Exadata

Doug Utzig and I will be presenting a session on planned maintenance for Exadata this year at Oracle Open World. The session is called 'Best Practices from Oracle Development for Patching and Upgrading Oracle Exadata'. If you are interested please enroll for session: CON8833 at Thursday, Oct 29, 10:45 a.m. in Moscone South 102.  


Oracle Exadata administrators often ask how to patch and upgrade Oracle Exadata components with zero downtime and minimum operational effort. Attend this deep-dive session to learn Oracle Maximum Availability Architecture best practices directly from Oracle Development for the upgrade of every software component of Oracle Exadata with no database downtime and maximum automation and ease. Learn how to easily determine when to upgrade and to what version, and how to achieve a predictable, successful upgrade experience. Topics for experienced administrators include how to properly maintain and upgrade a customized Oracle Exadata system and how deploying Oracle VM on Oracle Exadata affects software maintenance.

Rene Kundersma 

Monday Sep 08, 2014

OOW 2014 Session: CON7770 Best Practices: Consolidate with Oracle Exadata and Manage Resources and Availability

Tuesday September 30th at 5pm, together with Sue Lee (Oracle Server Technologies Director of Development), I will be presenting an interesting session on best practices for managing resources and high availability when consolidating on Oracle Exadata. This session is highly recommended for those planning to consolidate or already consolidating on Exadata. The session will cover planning, HA, and of course resource management aspects.

CON7770 Best Practices: Consolidate with Oracle Exadata and Manage Resources and Availability 

 René Kundersma 

Tuesday Apr 22, 2014

HA Best Practices for Database Consolidation

Read this new white paper to learn how to deploy and optimize Database-as-a-Service using Oracle Multitenant and Oracle MAA.

Monday Mar 10, 2014

Prereq Dependency Check Added for Exadata DB Server Updating

With this blog update some background on new functionality added to the Exadata utility. 

The recently added 'Prereq Dependency Check' feature in eliminates possible unexpected package dependency problems that occur during update by detecting dependency issues during the prerequisite check phase that could occur when the update is performed. (Linux package dependency issues sometimes happen after operators customizing the db server by installing additional software packages or newer releases of packages). The recommended use of this feature is to run with the –v flag (to verify prereqs only) days before the scheduled maintenance. will report dependency problems that must be resolved before proceeding with the update.

Note that the dependency check functionality is also run by default prior to performing the update. will now also list what packages will be removed for your update.

Some details:
  • Updates starting from - to any release earlier than
    • Dependency check is validated against 'standard' dependencies.
  • Updates starting from - to any release equal to or later than
    • Dependency check is first validated against 'exact' RPM dependency.
    • If 'exact' RPM dependency check passes it is assumed 'minimum' RPM dependency check will also pass.
    • If 'exact' RPM dependency check fails then 'minimum' RPM dependency check is run.
  • Updates starting from release do not have Prereq Dependency Check functionality.
  • will report what checks were executed for your update and which of them did 'pass' or 'fail'
    • If the dependency check is executed as part of –u and only 'minimum' RPM dependency check passes, then the new target will implicitly be changed to 'minimum' (which is equal to -m). 
    • If the dependency check is executed as part of –u and both 'exact' and ’minimum’ RPM dependency checks fail, then the operator will not be able to proceed with the update. For dependency checks that fail a separate report is generated.This report highlights to the failing package. The operator can then decide to either remove/install/update the failing package depending on what works best for that particular server.


  • Prereq run here -this is a prereq only run. Notice the ''Exact' package dependency check failed' and the ''Minimum' package dependency check succeeded'
    • Failing dependencies here  - for more details on what package cause the problem and what can be done to resolve it.
  • Update scenario here - see the same dependency checks and notice 'Update target switched to 'minimum''
Existing backups of the current image overwritten by default:

  • Existing backups of the current image on the inactive lvm will be overwritten by default. You can decide to skip (and retain) the backup by using the "-n" flag.

 Rene Kundersma 

Thursday Jan 16, 2014

Updating database servers using - part 3

When running for updating a db server the "-l" argument is always required to specify the source location of the new release/update. From here we will now call this  'location' the repository'.  The repository comes in two different flavors: as an ISO and as Oracle ULN channel. Let's start with the ULN channel.  

Exadata ULN Channels

For the different Exadata releases starting a corresponding 'base channel' is available on (ULN). Exadata customers can register to ULN with their CSI, subscribe to the channel (e.g. patch/release) they require and then synchronize one or more channels to with the local (in house) repository. This local repository can then be made available to the internal data-center by publishing it via a web server. 

Note: it is not recommended to use an Exadata db server as local YUM repository

Instructions how to subscribe a system to ULN and synchronize to a local repository are the same as for Oracle Linux and Oracle VM, so generic instructions can be found on OTN here. The README of a specific Exadata release will always mention what channels are made available for that release. You can also find this information via MOS 888828.1 and also in 1473002.1.

Additional to the 'base channel', there is also a 'latest channel'. Currently for Exadata there is a latest channel for 11.2 and 12.1 releases. The content of the 'latest' channel will never remain the same (unlike the 'base' channel) as long as there will be updates for that 11.2 or 12.1 release. When for example a new Exadata 12 release will be published this will be added to the existing latest channel (in addition to a 'base channel' also made available). This is the primary reason for the 'latest' channel being much larger (and taking more time to synchonize) than a base channel.

For Exadata installations on release later than, the 'latest' channel brings additional options. With the latest channel it's now possible to specify what release you want to update to. For example when on planning to update to a later (but not the latest 12.1.release, just as an example) you can use the 'latest' channel and specify the "-t" flag to specify what release you want to update to. 

Note that this can only be done with the 'latest' channel and that without specifying the "-t" argument by default the db server will be updated to the most recent release it can find in that channel. Of course there is also the option to just grab the 'base' channel and update without specifying any "-t' option.


  • updating with a latest channel specifying no argument (latest release in the channel will be used) here
  • updating with the latest channel to a specific release that not exists (a typo) here
  • updating to a specific release here

Exadata channel as ISO 

For those not able or willing to synchronize repositories with Oracle ULN there is also an ISO image available. The ISO image is built (and zipped) by Oracle and is only available for  'base' channel content. An ISO is ~1GB and about the same size as the sum of all packages of the corresponding base channel on ULN.

Using ISO or ULN

From an 'update' perspective there isn't much difference between using ISO or a http repository, only the location (-l) changes:

For local YUM repositories (synchronized from Oracle ULN):

./ -u -l http://myrepo/yum/unknown/EXADATA/dbserver/

For using an ISO (example with the iso):

./ -u -l ./ 

The ISO file should not be unzipped and there is no need to make an local 'loop mount' to use the iso - this is all done by the script


For each type of repository some validation checks will be done to see if it a usable is a repository, checks are done for expected files and also if the available Exadata release in the repository is a later release than the one currently installed - because if not, an update would not be possible. 

When specifying an http repository it's required to specify the top level directory containing the 'YUM metadata repository directory'. Typically this is the directory that has the 'repodir' directory in it. (see example here). When an http location cannot be identified as a valid repository an you would see a suggestion how to locate the right url.

Rene Kundersma

Wednesday Jan 15, 2014

Updating database servers using - part 2

Within the Oracle Exadata Database Machine documentation and README's you will generally find two types of backups for database server OS backup:
  • The Exadata Database Machine Owners Guide (Chapter 7) has instructions to create backups stored outside of the dbserver, for example on an NFS mount (see 'Creating a Snapshot-Based Backup of Oracle Linux Database Server')
  • - which creates a local copy of your active lvm.

In this post I will explain background and usage for both backups and how they integrate with

For backing-up and rolling-back Exadata dbserver OS updates  the script is used by For each upgrade by default the script is executed. When executed (either manually or via dbnodeupdate), the script creates a small snapshot of the 'active' sys lvm. The active sys lvm is the primary lvm that your current OS image is running on. For example:

[root@mynode ~]# imageinfo

Kernel version: 2.6.39-400.126.1.el5uek #1 SMP Fri Sep 20 10:54:38 PDT 2013 x86_64
Image version:
Image activated: 2014-01-13 13:20:52 -0700
Image status: success
System partition on device: /dev/mapper/VGExaDb-LVDbSys2

In the above example the active lvm is /dev/mapper/VGExaDb-LVDbSys2.The snapshot is created to have a 'consistent' 'view' of the root filesystem while the backup is made. After the snapshot is created, it's mounted by the same script and then it's contents are copied over to the inactive lvm. For lvm enabled systems, there are always 2 'sys' lvm's "VGExaDb-LVDbSys1" and "VGExaDb-LVDbSys2". VGExaDb-LVDbSys2 will automatically be created (on lvm enabled system) if not existing yet. For the example above, the 'inactive' lvm will be VGExaDb-LVDbSys1

Now, depending on how many files there are in the root (/) filesystem (based on your active sys lvm) the backup times may vary. Previous Grid and Database home installation zip files in /opt/oracle.SupportTools/onecommand will make the backup take longer (not the restore, which I will explain why later). Same for those who have many small files (like mail messages in /var/spool) - the backup may take longer. 

One of the first steps the script will doing when executed is making a backup with this script. Now, if you want to shorten your downtime and make this backup before the start of your 'planned maintenance window' you have 2 options: Either execute the script yourself or use with the "-b" flag to make a backup only before hand.

Example making a backup with here (see 'Backup only' for 'Action')

When you then have the downtime for planned maintenance and already have the backup you can then let dbnodeupdate skip the backup using the "-n" flag.

Example skipping a backup with here (See 'Create a backup: No')

Both Sys lvm's are 30GB each. The snapshot that will be created is ~1GB. It is recommended to keep this in mind when claiming the free space in the volume group to make your /u01 filesystem as big as possible. (the script checks for 2 GB free space in the volume group)

Now, when the update proceeds, the current active lvm will remain the active lvm. This is different than what happens on the cells where the active lvm becomes inactive with an update.  Typically you will only switch active sys lvm's when a rollback needs to be done on a db server, for example, an upgrade from to needs to be rolled-back. What happens then is nothing more than 'switching' the filesystem label of the sys lvm's, updating grub (the bootloader) and restoring the /boot directory (backed up earlier also by Then, a next boot will now have the previous inactive lvm as active lvm.

Rolling back with as in the example here (a rollback from to 

After booting the node, it's recommended to run again with the "-c" flag to relink the oracle home's again.


  • It's important, to make a new backup before attempting a new update.
  • In the above example, there is only talk about the sys lvm's. This means custom partitions including /u01 are not backed up. For regular node updates this is enough to rollback to a previous release but it's recommended to also have a backup of other filesystems inline to your requirements.
  • Nodes deployed without lvm will not have this option available
  • Rolling back db servers to previous Exadata releases with this procedure does not rollback the firmware
Backup / restore procedure owners guide chapter 7 

The backup made with the procedure in chapter 7 of the Oracle Exadatabase Database owners guide covers total node recovery.  Like the procedure a snapshot is used for a consistent view, then in this scenario a copy is placed outside of the db server (via NFS in this example).  This procedure allow you to backup every filesystem you require. In case of emergency - such as a non-bootable system, the node can be booted with the diagnostic iso. For non-customized partitions an interactive script will then question you to provide backup details and recover the node completely. For customized partitions steps (which are almost the same) can also be found in the owners guide.

Advantages /  Disadvantages

Both type of backups serve another goal. Also, these are just examples - of course customized backup and restore scenario's are also possible.The procedure as described in the owners guide requires external storage, while the script uses space on the node - but that is also where the risk is. The backup made with works well for the purpose of rolling back upgrades. With the automation of rollbacks can be done simple and quickly.

However - loss of critical partitions and/or filesystems will not be covered with this type of backup - so you may want to combine both types of OS backup. The general recommendation is to use the default built-in backup procedure when running dbnodeupdate to make easy rollback possible. But also backup the entire OS and customized filesystems outside of the database server with an interval based on your own requirements.

Rene Kundersma

Tuesday Jan 14, 2014

Updating database servers using - part 1

In this and future posts I am planning to describe some new functionality and background of starting with Oracle Exadata Database Machine release Some of this functionality will be directly available to the operator via the interface and can actually be used via an argument, however, some of the recent changes are made to make patching even easier, reduce human error and downtime.

You may also find some of the 'new features' described in MOS 1553103.1 'Exadata Database Server Patching using the DB Node Update Utility'

Exclusion/Obsolete list 

With updates to Exadata or later some packages on the database server will become obsolete. When updating a db server the script will mention an 'RPM exclusion list' and an 'RPM obsolete list' in it's confirmation screen. The 'RPM obsolete list' will list the all packages that will be removed by default during the update to (or later) when no action is taken by the operator.

As an example - click here

If you would like to find out first what obsolete packages will be removed you have to choose 'n' when prompted to 'Continue ? [Y/n]'. This will stop your current patching session. Then look at the contents of the freshly created 'obsolete list' (it should have the runid of your dbnodeupdate session in it's header). Example here

All the packages listed in the '/etc/exadata/yum/obsolete.lst' file will be removed by default - this has multiple reasons, mainly this is because these packages are not required anymore for Exadata functioning or they are considered a security risk. In case you would like to keep for example the 'java' package, you should create an 'exclusion file'  which is '/etc/exadata/yum/exclusion.lst' and put the 'java*openjdk' rpm name (or wildcard) in it.


[root@mynode u01]# cat /etc/exadata/yum/exclusion.lst

When  is restarted you would see that the 'exclusion file' is detected (an example here).

All packages you have put in the 'exclusion file' will still be listed in the obsolete file, but will not be removed when the confirmation screen says 'RPM exclusion list: In use (rpms listed in /etc/exadata/yum/exclusion.lst)'  with the update to or later.

Frequent releases of - keep an eye on it: and it's MOS note were designed / made to be released frequently and quickly when needed.This way can provide workarounds for known issues, emergency fixes, new features and best practices to the operator relatively quick.This reduces risk of people not keeping up to date with 'known issues' or not being sure it applies to them. Basically the same idea as with the patchmgr plugins.

Also, unlike to the storage servers some customization can be done on the db servers - best practices and lessons learned from this in regards to patching may also benefit your environment. For those engineers involved with upgrading Exadata database nodes, I'd like to emphasis to always check for the most recent release of when doing a db server update. I have already seen some people watching the releases closely, which is definitely a good thing.

Rene Kundersma

Thursday Dec 12, 2013

Enhanced lights-out patching for Exadata Storage Cells

The recently released Oracle Exadata version comes with multiple enhancements in the patchmgr utility. For those who don't know what the patchmgr utility is: the patchmgr utility is a tool Exadata Database Administrators use to apply (or rollback) an update to the Oracle Exadata Storage Cells. The enhanced patchmgr will have an option to send the operator an email for the most significant patch and rollback state changes. This eliminates the need to monitor the screen while the update or rollback is in progress.

In order to send the email you need to specify values for the '-smtp_from' and '-smtp_to' flags. Example as follows:

./patchmgr -cells ~/cell_group -patch -rolling \
           -smtp_from  \

Patchmgr will use the sendmail package which  is installed by default on the Exadata Database Server. It will start sendmail if it's not already started and assumes it's configured to deliver email for the domains you specify. You will recognize the format of the alerts patchmgr sends as they have the same formatting as the ASR emails.

The email you will receive when you enable this option can have the following end states for a cell: 

  • Started
  • Start Failed
  • Waiting
  • Patching
  • Rolling Back
  • Failed
  • Succeeded. 
  • Not Attempted 

The majority of the above states probably is self explanatory, however explaining 'Not Attempted' may help: 'Not Attempted' will be the final state of a cell when patching or rolling back (in a rolling fashion) of the current cell has failed. In that case the patching will stop and the remaining cells (which are not touched yet) are in the state 'Not Attempted'. So for example: imagine you are patching cel1,cel2 and cel3. When cel1 fails patching, then the end state for cel2 and cel3 will be 'Not Attempted'.

The following example will give you the idea: Patching starts in a rolling fashion from to for cel1,cel2 and cel3. 

Note: this is an example of the type of state changes you will see, actual patch timings are not correct and not relevant for this example. Also actual states or formatting may change in your release.

1. patchmgr was started in a rolling fashion. Initial state for all cells is 'Started'. This should be seen as this initialization phase. You will receive an email as follows:

2. The actual patching has begun since this is in rolling fashion, patchmgr starts with the first cel listed, which is cel1. The other two cells are waiting until cel1 finishes.

3. After some time patching of cel1 has completed. You will receive another status update in your inbox stating cel1 was updated successfully.

4. patchmgr continues with the next cell.

5. When succeeded (or failed) you will receive a state update via mail.

6. Last cell to be patched is cel3:

7. The last email you will receive will give an update stating patching was completed and lists the end state for all cells.

In case patching or rolling back would fail, you would receive a pointer to a release specific MOS note where you may find additional information.

Rene Kundersma 

Tuesday Oct 15, 2013

Database as a Service Online Forum

Join this online forum to hear from analysts and experts on how companies are beginning to transform with DBaaS, and learn the prescriptive steps your organization can take to design, deploy, and deliver DBaaS today.


Monday, October 21, 2013
9 a.m.–12 p.m PT / 12 p.m.–3 p.m. ET

Tuesday Sep 03, 2013

Exadata Database Machine Grid Infrastructure and Database Upgrade

Today on My Oracle Support two notes are made available for upgrading your 11.2 Grid Infrastructure and Databases running on the Oracle Exadata Database Machine to 

  • The first note is 1565291.1 This note has the steps for upgrading Grid Infrastructure and Database on BP12 and later to". 
  • The other note is 1555036.1 and has the steps for upgrading Grid Infrastructure and Database on BP11 and earlier to".

Rene Kundersma

Monday Aug 26, 2013

OOW2013: Consolidation and Virtualization on Oracle Exadata: How, What, Where, and When

Together with two colleagues from the MAA team I will be doing a presentation on consolidation for Oracle Exadata:

Title: Consolidation and Virtualization on Oracle Exadata: How, What, Where, and When (CON8530)

Abstract: How is database consolidation done on Oracle Exadata? This technical session looks at the past, present, and future consolidation solutions on Oracle Exadata. It describes all the consolidation options on Oracle Exadata, when to choose which, and the best practices for deploying an Oracle Database instance with each one 

When:  Monday, Sep 23, 3:15 PM - 4:15 PM - Westin San Francisco - Metropolitan I  

Link: here

Rene Kundersma 

Sunday May 26, 2013

Exadata Database Server Patching using the DB Node Update Utility

Today the 'DB Node Update Utility' ( utility) has been made available. The 'DB Node Update Utility' automates all the steps and checks to upgrade Oracle Exadata database servers to a new Exadata release and replaces the manual steps. The utility includes the latest best practices and workarounds for known issues. Where appropriate, the script uses existing 'One-Time Helper scripts' or 'yum update commands'. The script uses the 'build-in' '' script to perform backups before patching where possible. 

Four typical use cases for the utility are :

  • One-Time Setup  (updating procedure for Exadata releases running Oracle Linux 5.5 or later)
  • Updating database servers running Exadata releases later than on Oracle Linux 5.5 or later
  • Rolling back updates
  • Post-Patching (or Post-Rollback steps) (relinking the Oracle homes, enabling Grid Infrastructure to start)

For more details, demo's and downloads see My Oracle Support Note 1553103.1

Rene Kundersma

Wednesday Apr 10, 2013

Exadata Maximum Availability Tests

Recently Oracle published the below video created by our group (MAA  team) which demonstrates Exadata's High Availability in the face of hardware and software failures.  This video is a great showcase on how Exadata handles hardware and software failures. These are tests Oracle does on a daily basis for Engineered systems so every Exadata customer benefits  !

Exadata Maximum Availability Tests from ESG Media on Vimeo.

Tuesday Feb 19, 2013

Oracle Certification: Oracle Exadata Database Machine Administration

For those working on Exadata: Oracle University recently released a new test called "Oracle Exadata Database Machine Administration". This OU exam is very relevant for those designing and administrating Exadata environments.

Passing this test would give you the "Oracle Certified Expert" credential and prove you are able to work out the following key area's:
  • Architecture
  • Exadata Key Capabilities
  • Initial Configuration
  • Storage Server Configuration
  • I/O Resource Management
  • Consolidation
  • Migration
  • Bulk Data Loading
  • Monitoring & ASR
  • Backup and Recovery
  • Planned Maintenance
  • QOS

The test has been developed by Oracle Universities best instructors and I would really recommend taking the test if this is your field of expertise. It's a good opportunity to prove your skills. And for those looking for skilled Exadata specialists, I would recommend asking for people who passed this OU test additional to traditional 11g OCP and some basic Linux & Networking skills.

See: Oracle Certification: Oracle Exadata Database Machine Administration, Software Rel. 11.x

The above link also has references to recommend Exam preparation

Rene Kundersma

Thursday Oct 25, 2012

Demo on Data Guard Protection From Lost-Write Corruption

Today I received the news a new demo has been made available on OTN for Data Guard protection from lost-write corruption. Since this is a typical MAA solution and a very nice demo I decided to mention this great feature also in this blog even while it's a recommended best practice for some time.

When lost writes occur an I/O subsystem acknowledges the completion of the block write even though the write I/O did not occur in the persistent storage. On a subsequent block read on the primary database, the I/O subsystem returns the stale version of the data block, which might be used to update other blocks of the database, thereby corrupting it.  Lost writes can occur after an OS or storage device driver failure, faulty host bus adapters, disk controller failures and volume manager errors.

In the demo a data block lost write occurs when an I/O subsystem acknowledges the completion of the block write, while in fact the write did not occur in the persistent storage. When a primary database lost write corruption is detected by a Data Guard physical standby database, Redo Apply (MRP) will stop and the standby will signal an ORA-752 error to explicitly indicate a primary lost write has occurred (preventing corruption from spreading to the standby database).


  • MOS (1302539.1). "Best Practices for Corruption Detection, Prevention, and Automatic Repair - in a Data Guard Configuration"
  • Demo
  • MAA Best Practices

Rene Kundersma


Blog of Rene Kundersma, Consulting Member of Technical Staff at Oracle Development USA. I am designing and evaluating solutions and best practices around database MAA focused on Exadata. This involves HA, backup/recovery, migration and database consolidation and upgrades on Exadata. Opinions are my own and not necessarily those of Oracle Corporation. See


« November 2015