This letter studies distributed opportunistic channel access in a wireless network with decode-and-forward relays. All the sources use channel contention to get transmission opportunity. If a source wins the contention, the channel state information in the first-hop channel (from the source to its relay) is estimated, and a decision is made for the winner source to either give up the transmission opportunity and let all sources start a new contention, or transmit to the relay. Once the relay gets the traffic, it may have a sequence of probings of the second-hop channel (from the relay to the destination). After each probing, if the second-hop channel is good enough, the relay transmits to the destination and completes the transmission process of the source; otherwise, the relay decides either to give up and let all sources start a new contention, or to continue to probe the second-hop channel. The optimal decision strategies for the two hops are derived in this letter. The first-hop strategy is a pure-threshold strategy, i.e., when the first-hop channel signal-to-noise ratio (SNR) is more than a threshold, the winner source should transmit to the relay, and subsequently the second-hop strategy should let the relay keep probing the second-hop channel until a good enough second-hop channel is observed. Simulation results show that our scheme is beneficial when the second-hop channels have larger average SNR.