Re: Scan Listener

From: Guillermo Alan Bort <>
Date: Fri, 4 Oct 2013 23:59:48 -0300
Message-ID: <>

yes, you are correct. While SCAN Listener relocation is rather fast, you would have a window of a few seconds while the listener on an active node starts up, during which new connections using SCAN won't work. I highly recommend using at least two SCAN listeners to avoid just that. If you are working on a tiered environment and only an app server connects to the database, then the issue becomes a bit more complicated, depending on whether you have TAF or not, but as a rule if a database node goes down, then the app server will likely issue new connections to the database to make up for the lost ones (depending on how the app server is configured) and these new connections would of course fail because they would happen just as the node went down and while the SCAN listener is coming up on another node. Depending on the app server this could cause some issues.

Client/Server probably apps won't be affected as much because end users would have to actually attempt to connect in that tiny window. It is still possible, depending on how the app is built, that as soon as the connection dies, the user will attempt to reconnect, but in my experience with client/server apps, they usually won't notice the dead connection until they try to "save" or get some information, and only then the app will either error out, or attempt to connect again (most likely it will error out and have the user connect again). Then again, if you have the resources to have a RAC database, you probably have most apps in an app server.

sorry if I digress a little.. it's been a while since I got to exercise the "what if" part of my brain :P



On Fri, Oct 4, 2013 at 10:58 PM, Sanjay Mishra <> wrote:

> Hi
> I want to check that if any RAC environment has used one DNS based Scan
> listener(DBA not provided correct requirement to the network and they
> assigned only one IP) and so RAC install has only configured one SCAN
> listener. Now if this is 4 Node RAC, and if the node containing SCAN
> listener goes down or crashed, it will take time to bring this scan
> listener on other surviving node. So want to check if during this time any
> app trying to connect using SCAN listener, then isn't this cause outage
> until SCAN listener is not up on second node. Also how long it will take to
> bring SCAN listener on surviving node. I know we can add more scan listener
> now but want to check the concept in the above scenario.
> Sanjay
> --

Received on Sat Oct 05 2013 - 04:59:48 CEST

Original text of this message