Hello,

We have an Oracle Database cluster running as a RAC on 2 servers. Currently we have only managed to get Oracle v2 SQL sensors to work when connecting directly to one database node and using the SID of the database instance running on that specific node.

We'd like to use the Oracle scan address or similar so that the database can load balance requests and failover properly should there be a node failure.

Do you have any information on best practise when connecting to Oracle Databases using RAC or scan addresses?

Any information greatly appreciated

Regards,

Tom

Article Comments

This is unfortunately something that we do not have much experience with and have not tested in-house. What issues are you running into when you are trying to use the sensor with the RAC or scan addresses?


May, 2015 - Permalink

Hi Greg,

Thanks for the reply. The error we receive is a fairly clear one:

"ORA-12505: TNS:listener does not currently know of SID given in connect descriptor"

The device DNS/IP is set to the scan address (e.g. "database-scan") rather than the DNS for the individual database nodes (e.g. "node1" or "node2"). Each node has a separate instance of the database running, "database1" and "database2" respectively, but the scan address exposes a database service called simply "database", and the scan address directs the connection request onto a particular node.

These details work with other direct-connect methods when connecting to the database.

I notice the Oracle SQL sensor (not the v2 one) appears to support an external Oracle client - this may be our answer.

I also notice in the manual page the Oracle SQL v2 sensor does explicitly state it doesn't support RAC, so I don't think we'll get it working with this sensor type.

Thanks,

Tom


May, 2015 - Permalink

Ok, please let us know if the older sensor works.


May, 2015 - Permalink

Hi,

For anyone who is interested, an update to PRTG seems to now allow this functionality.

The "new" Oracle v2 sensor has an option to connect to either SID or SERVICE_NAME, whereas previously it only supported SID.

I can now connect to a Service running on the Database via the oracle scan address. This means we can use one device for a multi-node database, and the queries will be distributed across the database nodes rather than connecting to a single node. It also means if a database node is unavailable, the monitoring is not impacted (we have separate sensors to detect if a specific node becomes unavailable)

Thanks to Paessler for the update!

Please note we are running 15.x.19, however I am not 100% when this update actually appeared

Regards, Tom


Nov, 2015 - Permalink