Grouper Users - Open Discussion List

[Nathan Kopp <>]

Note that if you use “TYPE=select” in the CONNECT_DATA section of your Oracle RAC connection string, the Oracle JDBC client itself will automatically retry “select” statements when it fails over internally.  (See my previous message on this topic for more details and a link with more documentation.)

 

-Nathan

 

From: [mailto:] On Behalf Of Chris Hyzer
Sent: Thursday, April 26, 2012 11:35 AM
To: Gagné Sébastien;
Subject: [grouper-users] RE: Database high availability with Oracle RAC and SCAN

 

Grouper does not retry queries if there are errors…  that seems like a complex feature for a non-idempotent transactional system (e.g. long running transactions, inserts, autonomous transactions, etc)…

 

The grouper client will do this in 2.1, though it is idempotent, you cant have multiple operations in one transaction, so it is easier… J

 

Thanks,

Chris

 

From: [mailto:] On Behalf Of Gagné Sébastien
Sent: Thursday, April 26, 2012 9:17 AM
To:
Subject: [grouper-users] Database high availability with Oracle RAC and SCAN

 

Hello,

We are deploying Grouper in an high availability environment using an Oracle database and we would like to confirm Grouper’s behaviour regarding databases.

 

We have 2 database servers using RAC mode (real application cluster) behind a SCAN address. Basically applications configures their DB URL using the single URL (see grouper config example below) and the SCAN+driver takes cares of getting the IP address of an available server (usually based on load or other configuration). I successfully used Grouper with the RAC+SCAN database architecture, but we were wondering if it will react properly if/when one of the servers becomes unresponsive.

 

We would like to know how Grouper reacts regarding database failures.

Will Grouper automatically retry failed commands ? If I does, the SCAN will give the IP of the other available server.

Will grouper report an error and stop working ? Then high availability of the database is non-existent.

 

Grouper connect string example :

jdbc:oracle:thin:@//db-alias.umontreal.ca:1521/SERVICENAME.UMONTREAL.CA

 

In comparison, when installing Oracle products we are using a connect URL like :

db-alias.umontreal.ca:1521^db-alias.umontreal.ca:

 

This format enables RAC mode, meaning the product will try the second URL if the first one fails. Since SCAN is a single URL we have it 2 times to force the application to retry the same connection, but this time the SCAN will give another IP for the database server.

 

Thank you

 

 

Sébastien Gagné,     | Analyste en informatique

514-343-6111 x33844  | Université de Montréal,

                     | Pavillon Roger-Gaudry, local X-100-11