Abstract: A computer network comprises first and second computers, each having a recovery facility for recovering a two-phase commit procedure and a recovery log. The log comprises a log name log which stores the names of other logs involved in a commit scope and a sync point log which stores sync point phase information. The log names are exchanged prior to a request for sync point. If a failure occurs during sync point, each recovery facility in the commit scope again exchanges log names to verify that there was not a log failure since the prior exchange. Absent a log failure, the logs are used to recover from the sync point failure. Communication facilities are interposed between the first and second computers and include log name exchange tables which record the previous exchanges of log names between the first and second computers.

Abstract: In a computer system or process, sync point managers are distributed throughout each real machine for a plurality of execution environments, but all of the execution environments and sync point managers within one real machine share a common recovery facility and recovery log. A common recovery log is used by the recovery facility for all of the execution environments in the system. Different systems are interconnected by a communication facility and each has its own recovery facility and recovery log. A protected conversation can be initiated between the first and second execution environments in the same real machine, and the sync point managers within the respective execution environments coordinate the two-phase commit procedures associated with the protected conversations. A conversation manager within each real machine assists in routing the conversation between the first and second execution environments.

Abstract: To recover a failed commit procedure, a first recovery facility collects and compares sync point state information from a variety of different types of resources including a data repository and a protected conversation. The comparison of the sync point state information is made automatically in response to a sync point failure event. A second recovery facility serving another execution environment notifies the first recovery facility after a failure and subsequent recovery procedure of the second execution environment has completed. The first recovery facility makes automatic cyclic attempts to communicate with the second recovery facility. The first recovery facility can receive a manually initiated but automatically simulated response of sync point state information when there is a failure such that actual sync point state information cannot be received.