I am honored to have a blog post on the Pivotal Blog!
5.0.6 now uses UTC for session management rather than the local timezone. This handles the situation where the database is in a different timezone than the host where Outsourcer is running.
5.0.6 is also a stable release. Anyone still using 4.x should upgrade to 5.0.6.
I have made improvements around the starting and stopping of gpfdist processes. It now handles the situation where a gpfdist process doesn’t start properly and instead of failing, it will pick the next available port. I also completely separated osstart and uistart. The User Interface does not need to be running in order for Outsourcer to load tables in the Queue or select from Custom SQL Tables.
This version is a release candidate as testing has shown it is very stable.
Here is another new release of Outsourcer 5. I will likely have another release by the end of the week to further clean up the code so be on the lookout for that. I will update the documentation for release 5.0.5 as well.
I am making changes to Outsourcer 5 due to the move from External Web Table to External Tables using gpfdist. In my testing with 4 concurrent jobs, gpfdist seemed to work fine with only one process handling the load. After stress testing, I have learned that gpfdist is not multi-threaded and a single process can not handle the load of multiple jobs.
I have resolved this issue in 5.0.3 by starting and stopping gpfdist for every job. Upon further stress testing, I identified a potential problem when gpfdist fails to start properly. This would cause multiple jobs to fail in a cascading fashion.
5.0.4 resolves the scenario when gpfdist fails to start. The single job will fail and not affect other jobs.
This release also decouples the User Interface from Outsourcer processing jobs in the Queue. You can run Outsourcer without the User Interface running at all.
I also identified a scenario where Outsourcer 5 is installed on the Master was causing problems. The installer was expecting the installation to happen on the Standby-Master or on a dedicated ETL server so the installer would create a new .pgpass file and it would also backup an existing one. This would remove the gpmon entry for and prevent Command Center from working correctly. 5.0.4 now keeps the gpmon entry, if found, in the .pgpass file. If you have a Standby-Master or an ETL host, install Outsourcer 5 there instead.
And lastly, if you just want a stable release of Outsourcer that isn’t changing frequently, use Outsourcer 4.1.6 which hasn’t changed since January and uses External Web Tables.
Outsourcer 5.0 is now available. This version addresses most of the requests that I have received since the release of 4.0. I’m very excited to complete this work!
Outsourcer now uses External Tables with gpfdist rather than External Web Tables. This is a big change and allows you to install Outsourcer on any host that is accessible to the Greenplum Segment Hosts / HAWQ Data Nodes. It also reduces the load on the Master server.
Installation does more checking and automation.
Root is no longer needed for installation.
start_all and stop_all scripts are provided to ease starting and stopping Outsourcer.
A new set of screens for creating your own Custom External Table with a pre-defined SQL Statement.
You can install multiple copies of Outsourcer to load data to multiple databases.
1. Error messages raised from Oracle and SQL Server could include special characters which would cause the thread in Outsourcer to die and the status remain labeled as “processing”. Error messages no longer have escape and quote characters to prevent this problem.
2. On new installations, the permissions for the Oracle and SQL Server Jar files would remain owned by root. This has now been corrected to be the admin user (gpadmin).
3. On the “Sources” screen, the label for Append-Only now is correct for Greenplum Database 4.3 and reflects that these are really “Append-Optimized” tables.
1. Added External Tables to enable starting and stopping the UI rather than having to ssh to the Master host.
2. “Queue” screen now provides the ability to cancel processing jobs.
3. Stopping Outsourcer Queue Daemon (Environment Screen), now will cancel currently running Jobs in the Queue.
4. Starting Outsourcer Queue Daemon (Environment Screen), now will make sure orphaned jobs in the Queue will be marked as Failed and if any are currently running, it will cancel these jobs too.