Yes, this patch just adds retries for non-idempotent ops. The intent of this patch to recover from the downtime when a standby namenode is becoming the active one. Currently, if HBase tries to do a non-idempotent operation (like create a file), it fails because NN is unavailable. This usually results in aborting the HBase server. This patch tries to cover that by retrying the op after sometime.
FileSystem URI will be changed
The idea of HA support is that a client doesn't need to change any url whatsoever, otherwise its not really a failover.
Another, for HDFS HA, fs.defaultFS can be configured as "hdfs://mycluster, does this patch support "hbase.rootdir" configured "hdfs://mycluster/hbase ?
Not sure I follow. If you mean whether properties fs.defaultFS and hbase.rootdir are supported or not, then they already are. Please look at HRS and MFS classes. If you mean a change in their value because of failover, then it is not required.
Hope this helps.