This week on the podcast, Kyle shares a great MOS document that explains how psadmin issues tuxedo commands and asks if we could hook into that setup. Then Dan and Kyle discuss their ideal refresh process if they were to start from scratch.
This week on the podcast, Kyle and Dan revisit the ACM (Automated Config Management). We discuss what the ACM is, how ACM works with the DPK, building custom ACM plugins and how to share ACM plugins.
- Revisiting ACM @ 1:00
- Using ACM with the Search Framework @ 5:00
- ACM and the DPK @ 9:00
- Building ACM Plugins @ 13:30
- Plugins that are missing @ 21:45
- GitHub Repository to Share ACM Plug-ins @ 30:00
- IDEA – More options for “Copy to File”
This week on the podcast, Frank Dolezal joins us to talk about how he measure user testing and how it builds confidence when deploying updates. We also discuss self-service and fast database refreshes, and what skills to look for when hiring a PS Admin.
- Introducing Frank @ 1:00
- Creating Metrics for User Testing @ 4:00
- Getting Users to Test @ 11:00
- Justifying testing through metrics @ 17:00
- Data Sources to gather testing metrics @ 23:00
- Self Service Refreshes @ 28:30
- Planning Catch-up Projects @ 39:00
- Changing PT Release Schedule? @ 43:00
- Oracle said that there was no plans to change, so this is speculation only
- Skillsets when hiring a PS Admin @ 48:00
This week on the podcast, Kyle and Dan talk about Query Manager bugs, changing how they apply CPU Patches, and testing partial database refreshes. Then Dan gives an overview of Oracle Configuration Manager and laments it’s unsupported status.
- Collaborate Sessions
- PeopleTools Ideas Page @ 3:30
- IDDA Logging Follow-up @ 4:15
- Query Manager / JSON bug @ 6:00
- Changing how you apply CPU Patches @ 10:15
ProcessRestartMemoryLimitParameter @ 18:15
- PTF being slow @ 29:15
- SRID in Access Log @ 32:30
- Populating Correlation Fields @ 35:30
- Partial database refresh to grab old code @ 37:30
- [Keeping Older PI Images] (https://support.oracle.com/epmos/faces/DocumentDisplay?id=1579720.1) @ 40:30
- How do you do it: Booting domains: Serial or Parallel? @ 45:45
- Oracle Configuration Manager @ 48:15
A PeopleSoft refresh is one of the more common tasks for a PeopleSoft Administrator or DBA. There are many ways to accomplish this task, but it usually involves a database restore from backup, custom SQL refresh scripts and potentially ACM steps. Depending on the level of effort put into the refresh scripts, there can also be manual steps involved. This approach is tried and true, but tends to lack the speed and flexibility that we are starting expect with the delivery of the PeopleSoft Cloud Architecture toolset. Nightly or Ad-hoc refresh environments and quickly provisioned temporary environments are just a few use cases that would benefit greatly from refresh process improvements. I have been doing some exploring in this area recently and would like to share a few thoughts. First, a quick overview of some Oracle tools and features that I have been leveraging.
Oracle Data Guard is a tool that gives you high availability, data protection and disaster recovery for your databases. At a high level, it consists of a primary database and one or more standby databases. These standby databases are transactionally consistent copies of the primary database. Therefore, if the primary database goes down, the standby can be switched to primary and your application can keep on rolling.
Physical vs. Snapshot Standby
There are multiple types of standby databases that can be used with Data Guard. I’d like to briefly explain the difference between
Physical Standby and
Snapshot Standby. A physical standby is a database that is kept in sync with a primary database via
Redo Apply. The redo data is shipped from the primary and then applied to the physical standby. A snapshot standby is basically a physical standby that was converted to a snapshot, which is like a point in time clone of the primary. At this point we can use the snapshot to do development, testing, etc. When we are done with our snapshot, we can then convert it back to a physical standby and it will once again be in sync with the primary database. This is accomplished by taking a restore point when the snapshot conversion happens. The whole time the standby is in snapshot mode, the redo data is still being shipped from the primary. However, it is NOT being applied. Once we convert back to physical, the restore point is used to restore and then all waiting redo is applied.
With Oracle 12c, we have the introduction of multitenant architecture. This architecture consists of Container(CDB) and Pluggable(PDB) databases. This setup makes consolidating databases much more efficient. It also gives us the ability to clone a PDB very easily. Cloning a PDB between different CDBs can even be done via a database link. Having a true multitenant setup does require additional licensing, but you can have a CDB-PDB setup without this extra licensing cost if you use a single instance(Only one PDB per CDB). Here is a great video overview of multitenant.
Now that we have an idea of what these tools and features gain us, let’s think about how to put them to use with database refreshes. Both of these approaches assume the use of Data Guard and PDBs. Having a true multitenant setup would be most efficient but a single instance setup will work just fine. I would recommend you have a dedicated standby database for your refreshes, versus using the same standby you rely on for HA\DR. It would also make sense for the standby to be located on the same storage as the PDBs you will be refreshing. Neither of these are requirements, but I think you will see better performance and lessen the risk to your HA\DR plan.
The use case we will use for this example is a sandbox PeopleSoft database. This sandbox will be scheduled to refresh nightly, giving the business an environment to test and troubleshoot in with data from the day before. The refresh could also be run adhoc, if there is a need during the business day. So the goal is to have this fully automated and complete as fast as possible.
Clone Standby Approach
This approach will be to take a snapshot of our refresh standby database and clone it, overlaying our previous sandbox PDB. After this is completed, we will need to run custom SQL scripts or ACM steps to prepare the refreshed PDB. Finally, we will restore the refresh standby back to a physical standby database. This blog post by Franck Pachot gives a quick overview of the SQL commands needed to accomplish most of these steps.
- Convert the refresh source physical standby to a snapshot standby.
- Open the refresh source PDB as read only.
- Create database link between the sandbox target CDB and the refresh source PDB.
- Drop the sandbox target PDB and create a clone from the refresh source PDB.
- Open the new clone sandbox PDB.
- Cleanup the sandbox PDB.
- Check for errors.
- Patch the PDB to the patch level of the CDB, if needed.
- Run custom SQL scripts or ACM steps against sandbox PDB for PeopleSoft setup.
- Convert the refresh PDB back to physical standby.
Snapshot Standby Approach
This approach is somewhat similar, except we won’t be doing any cloning. Instead, we will be using the actual snapshot standby itself as our database. Since we know this sandbox database will be refreshed nightly, we can stay in snapshot standby mode all day and then switch to physical standby mode briefly at night, applying redo data to sync up with our primary production database. After that is done, we will then switch back to snapshot mode and run our custom SQL scripts and ACM steps. This will require a dedicated standby database and should only be used with a frequent refresh schedule. Since the redo data continues to ship during snapshot standby mode, the redo data will start to backup. The volume of this redo data backing up could become an issue if it gets too large, so you will need to do some analysis to make sure you can handle it based on your refresh interval.
- Create a sandbox PDB as a physical standby, with primary database being production.
- Convert sandbox to a snapshot standby.
- Run custom SQL scripts or ACM steps against sandbox for PeopleSoft setup.
- Use the snapshot standby sandbox PDB as your normal database; connecting app and batch domains, etc.
- Wait until next refresh interval.
- Convert sandbox from snapshot standby to physical standby.
- Restore point will be used and redo data applied, syncing up with current primary database state in production.
- Covert sandbox from physical standby to snapshot standby.
- Run custom SQL scripts or ACM steps against sandbox for PeopleSoft setup.
Those are just two ideas, but you can see that there are probably many variations of these approaches that will work. Leveraging Data Guard and PDBs really gives you many options to choose from. I have been using the Clone Standby approach recently and have packaged up the scripts, including bouncing app\batch domains, in Oracle Enterprise Manager as a job. This gives me push button refreshes with a turn around time under 20 minutes. I have been able to provide adhoc refreshes for emergency production troubleshooting to the business multiple times in just a few months since implementing this approach. This is a very powerful tool to have and is well worth the effort to get your refreshes fast, efficient and automated.