Have a look at the thread this post An I/O error has occurred while writing a response message entity to the container output stream - #39 by matt1 belongs to.
That post defines a filter to filter out the error messages as the issue was not resolved in that thread.
Most probably you get lots of the same error message being logged at the same time.
This is why the CPU is going up and disk is filled with the same error message again and again.
