Thanks Larry,
It turns out that the clock noise is flowing through another input of the circuit. I grounded that input and the glitch is greatly reduced.
The filter capacitor is also useful to eliminate the jumps when the DAC receive alternative input that differ by 1 LSB.