This　paper　is　concerned　with　the　consensus　problem　of　general　linear　dynamic　multiagent　systems　with　stochastic　sampling.　In　this　paper,　the　sampling　intervals　randomly　switch　between　two　different　values.　The　communication　topology　between　agents　is　fixed　and　directed.　Full-and　reduced-order　observers　are　designed　based　on　neighbor　agents＇　relative　output　information.　The　algorithms　to　construct　such　observers　are　also　provided.　By　using　the　estimated　states　of　the　agents,　the　observer-based　consensus　protocol　with　stochastic　sampling　are　presented.　Sufficient　conditions　to　ensure　consensus　in　mean　square　are　derived　by　using　Lyapunov　stability　theory.　Finally,　simulations　are　given　to　examine　the　effectiveness　of　the　proposed　methods.