To determine whether a notification has failed, notifications use the system clocks of the machines that host the clustered Integration Servers. Synchronizing the clocks of all machines in the cluster is critical for the proper execution of clustered polling notifications.

However, in time these clocks might become un-synchronized. Therefore you should anticipate the effect of “clock drift” when you establish values for the keepAliveExpireTimeout server-specific setting and Max Process Time notification-specific setting. Clock drift is the time difference between the clocks. As a guideline, add to the keepAliveExpireTimeout and the Max Process Time settings two times the maximum clock drift you anticipate.