Statistical Inference — type I error of a randomised Poisson test

To test mean of the Poisson distribution $H_0 : \lambda = 1$ versus $H_1 : \lambda = 2$, based on the single observed value from the Poisson distribution, the following procedure is considered: "Reject $H_0$ with probability one when the observed value $(x) > 1$. If $x = 1$ is observed, then toss an unbiased coin and reject the null hypothesis if head comes up." What is type I error in the experiment?