The install went very smoothly until the "First Run" script, which failed at this step:
Execute command SparkUploadJarServiceCommand on service Spark
Command (156) has failed:
Wed Nov 12 21:02:29 GMT 2014
Wed Nov 12 21:02:29 GMT 2014: Detected CDH_VERSION of 
Wed Nov 12 21:02:29 GMT 2014: Uploading Spark assembly jar to '/user/spark/share/lib/spark-assembly.jar' on CDH 5 cluster
14/11/12 21:02:34 WARN hdfs.DFSClient: DataStreamer Exception
org.apache.hadoop.ipc.RemoteException(java.io.IOException): File /user/spark/share/lib/spark-assembly.jar._COPYING_ could only be replicated to 0 nodes instead of minReplication (=1). There are 0 datanode(s) running and no node(s) are excluded in this operation.
So, at this point the installer is still doing its thing, and I haven't stopped/started any services.
Not many hits on Google for "SparkUploadJarServiceCommand" - but any issue I have found mentions ports. I'm running Ubuntu 14.04 LTS.
Any help you can give would be appreciated - can I "skip" the "First Run" script and manually kick off services?
The most probably cause of this would be that forward and reverse hostname resolution on all hosts of your cluster is not fully and correctly set up such that each host knows itself and all other cluster nodes properly in both resolution directions.
If you're still in the install wizard, open a separate tab to the Cloudera Manager main UI and go to the Hosts tab, then run the Host Inspector. If it comes back noting that any name resolution is not correct, resolve those matters before proceeding any further.
Let us know how you get on. Thanks!