How yarn memory is allocated in emr? How to increase yarn memory?

EMR cluster YARN memory is defined as per the configuration defined here. Instance typer5.8xlarge has 253952MB defined in yarn.nodemanager.resource.memory-mb ( considering HBASE isn't installed in the cluster), so the cluster's YARN memory becomes 253952 * 6 MB ( 4 core and 2 task ).

This is to kept in such a way that there are some resources left for the OS and other processes.

You can tune the parameter in the cluster using EMR reconfiguration if you wish but in general tuning this is not recommended.

For more YARN cluster, consider to use EMR scaling.