Completeness detection of monitored globally distributed synchronous and asynchronous transactions
View Patent ↗A method and system for detection of completeness of tracing data describing globally distributed, synchronous and asynchronous transaction in real time is presented. Monitored transactions may be executed on different computers systems, connected via computer networks with different and fluctuating latencies, and timing systems of the computers systems may be subject to different time drifts. The system and method may be implemented by different, multiple collector nodes responsible for collecting and forwarding transaction tracing data to a monitoring node, responsible for correlating transaction tracing data. Collector nodes and monitoring node may reside on different computer systems, which may be connected via computer networks with different and fluctuating latencies, and timing systems of the computers systems may be subject to different time drifts. Completeness detection may consider different time drifts and different, fluctuating network latencies of all involved computer system and computer networks to translate time information received with tracing data into a virtual timing system, which may be used to implement a deterministic algorithm to detect completeness of tracing data describing the execution of distributed, synchronous and asynchronous transactions.
1. A method comprising:
(A) selecting a start path record representing a start path of a distributed transaction, wherein the start path corresponds to execution of a first part of the distributed transaction, wherein a plurality of application nodes executes parts of the distributed transaction, wherein the plurality of application nodes execute on a plurality of computer systems, wherein a first application node in the plurality of application nodes executes a first part of the distributed transaction, wherein a first plurality of agents is installed on the plurality of application nodes, wherein at least one of the plurality of agents is installed on each of the plurality of application nodes, wherein the start path does not have a parent path;
(B) identifying a completion timestamp, represented in a timing system of a destination node executing on one of the plurality of computer systems, indicating a completion time of the start path;
(C) identifying an oldest latest event timestamp, represented in a timing system of the monitoring node, from among a plurality of latest event timestamps associated with a second plurality of agents, the second plurality of agents including the first plurality of agents that are installed on the plurality of application nodes;
(D) determining whether the oldest agent timestamp represents a time that is later in time than the completion timestamp;
(E) if the oldest agent timestamp is determined to represent a time that is later in time than the completion timestamp, then setting a completeness state of the start path to indicate that the start path has completed;
(F) if the oldest agent timestamp is determined to represent a time that is later in time than the completion timestamp, then determining whether all sub-paths of the start path have completed;
(G) if it is determined that all sub-paths of the start path have completed, then setting a completeness state of the start path to indicate that the start path has completed; and
(H) if it is not determined that all sub-paths of the start path have completed, then, if a timeout period associated with the start path has elapsed since the start path has initiated, then setting a completeness state of the start path to indicate that the start path was completed by timeout.
2. The method of claim 1 , wherein (E) further comprises setting completeness states of all descendant paths of the start path to indicate that the descendant paths have completed.
3. The method of claim 1 , wherein the start path includes at least one asynchronous part, and wherein (D) comprises determining whether the oldest agent timestamp represents a time that is later in time than a sum of the completion timestamp and an asynchronous wait time.
4. The method of claim 1 , further comprising:
(I) if the oldest agent timestamp is determined to represent a time that is later in time than a sum of the completion timestamp and an asynchronous wait time, then determining whether all sub-paths of the start path have completed; and
(J) if it is determined that all sub-paths of the start path have completed, then setting a completeness state of the start path to indicate that the start path has completed.
5. A system comprising at least one non-transitory computer-readable medium comprising computer-program instructions executable by at least one computer processor to perform a method, the method comprising:
(A) selecting a start path record representing a start path of a distributed transaction, wherein the start path corresponds to execution of a first part of the distributed transaction, wherein a plurality of application nodes executes parts of the distributed transaction, wherein the plurality of application nodes execute on a plurality of computer systems, wherein a first application node in the plurality of application nodes executes a first part of the distributed transaction, wherein a first plurality of agents is installed on the plurality of application nodes, wherein at least one of the plurality of agents is installed on each of the plurality of application nodes, wherein the start path does not have a parent path;
(B) identifying a completion timestamp, represented in a timing system of a destination node executing on one of the plurality of computer systems, indicating a completion time of the start path;
(C) identifying an oldest latest event timestamp, represented in a timing system of the monitoring node, from among a plurality of latest event timestamps associated with a second plurality of agents, the second plurality of agents including the first plurality of agents that are installed on the plurality of application nodes;
(D) determining whether the oldest agent timestamp represents a time that is later in time than the completion timestamp; and
(E) setting a completeness state of the start path to indicate that the start path has completed if the oldest agent timestamp is determined to represent a time that is later in time than the completion timestamp;
(F) determining whether all sub-paths of the start path have completed if the oldest agent timestamp is determined to represent a time that is later in time than the completion timestamp;
(G) setting a completeness state of the start path to indicate that the start path has completed if it is determined that all sub-paths of the start path have completed; and
(H) setting a completeness state of the start path to indicate that the start path was completed by timeout if: (1) it is not determined that all sub-paths of the start path have completed and (2) a timeout period associated with the start path has elapsed since the start path has initiated.
6. The system of claim 5 , wherein (E) further comprises setting completeness states of all descendant paths of the start path to indicate that the descendant paths have completed.
7. The system of claim 5 , wherein the start path includes at least one asynchronous part, and wherein (D) comprises determining whether the oldest agent timestamp represents a time that is later in time than a sum of the completion timestamp and an asynchronous wait time.
8. The system of claim 5 , wherein the method further comprises:
(I) determining whether all sub-paths of the start path have completed if the oldest agent timestamp is determined to represent a time that is later in time than a sum of the completion timestamp and an asynchronous wait time; and
(J) setting a completeness state of the start path to indicate that the start path has completed if it is determined that all sub-paths of the start path have completed.