OD weirdness

speck

We are having a very strange behavior

We have an OD script running that is moving files to 2 serverfarms with 2 servers in each farm, as well as a data deploy segment.

When we run the script without the datadeploy portion, it does deploy the file to all servers as expected and, exits with code 0

When we add the datadeploy the file appears to be delivered to all servers and the database is properly updated, but we get error code 2 as our response.

When we remove 1 server from each of the server farms, so that each of the 2 farms just contains 1 server each, it runs without issue with the data deploy and returns code 0.

Any ideas what's going on? And suggestions to fixes would be appreciated.

Adam Stoller

We are having a very strange behavior We have an OD script running that is moving files to 2 serverfarms with 2 servers in each farm, as well as a data deploy segment. When we run the script without the datadeploy portion, it does deploy the file to all servers as expected and, exits with code 0 When we add the datadeploy the file appears to be delivered to all servers and the database is properly updated, but we get error code 2 as our response. When we remove 1 server from each of the server farms, so that each of the 2 farms just contains 1 server each, it runs without issue with the data deploy and returns code 0. Any ideas what's going on? And suggestions to fixes would be appreciated.

This is a classic issue - when you run DD as a DNR as part of a fan-out deployment - the DNR will run for each leg of the fan-out deployment - and when you're deploying new data to the database it signifies the need to do an INSERT - but when the deployment is run multiple times, the first time is fine, but the subsequent attempts will fail because the INSERT won't work if the data is already present in the database.

You need to break up your farm and create two deployment definitions and two execDeployment's. One farm contains a single node and the execDeployment will run the dnrDeployment - the other farm can contain one-or-more nodes, and it's execDeployment does not run the dnrDeployment.

The alternative (which was described in another thread recently) is to try to use dnrDeploymentJob - the problem with this approach is that a failure of the DD part happens too late to cause a rollback of the OD part.