Methods and systems to manage alerts in a distributed computing system

1. A method stored in one or more data storage devices and executed using one or more processors of a computer system to manage alerts generated in response to log messages generated by one or more event sources that run in a distributed computing system, the method comprising: providing a graphical user interface that enables a user to define a dominant alert and select two or more subsumed alerts generated by the one or more event sources; searching log-message files of the one or more event sources for log messages that correspond to the two or more subsumed alerts; generating a single alert message that describes the dominant alert when a log message that corresponds to one of the two or more subsumed alerts is identified; and executing remedial action to correct a problem represented by the dominant alert.

2. The method of claim 1 wherein providing the graphical user interface comprises providing a graphical user interface that enables a user to select a duster of alerts for the one or more event sources, the duster of alerts comprises two or more alerts.

3. The method of claim 1 wherein providing the graphical user interface comprises providing a graphical user interface that enables a user to select the dominant alert and one or more subsumed alerts for the event sources, a cluster of two or more alerts comprises the dominant alert and the one or more subsumed alerts.

4. The method of claim 1 wherein searching the log-message files of the one or more event sources for log messages that correspond to the two or more subsumed alerts comprises: identifying log messages of one or more log-message files of the one or more event sources with time stamps in a selected time interval; determining the non-parametric tokens of each log message in the selected time interval; and comparing non-parametric tokens of each log message with keywords of each of the two or more subsumed alerts to detect log messages that correspond to the two or more subsumed alerts.

5. The method of claim 1 wherein searching the log-message files of the one or more event sources for log messages that correspond to the two or more subsumed alerts comprises: identifying log messages of one or more log-message foes of the one or more event sources with time stamps in a selected time interval; determining the non-parametric tokens of each log message in the selected time interval; comparing non-parametric tokens of each log message with keywords of each of the two or more subsumed alerts to detect a log message that corresponds to a dominant alert, the dominant alert is one of the selected two or more subsumed alerts and is the dominant alert; and comparing non-parametric tokens of each log message with keywords of each of the two or more subsumed alerts to detect one or more log messages that correspond to one or more subsumed alerts, the subsumed alerts are one or more of the two or more subsumed alerts.

6. The method of claim 1 wherein searching the log-message files of the one or more event sources that correspond to the two or more subsumed alerts comprises: identifying log messages of one or more log-message files of the one or more event sources; determining the non-parametric tokens of each log message: comparing non-parametric tokens with keywords that identify an alert; when a log message contains an alert, incrementing an alert count associated with the alert; and for each pair of alerts co-occurring in a sliding time window applied to the one or more log-message files, computing a relative alert count that represents a count of the number of times a pair of alerts are triggered in time limits of the sliding time window; and for each pair of alerts co-occurring in a sliding time window applied to the one or more log-message files, computing a relative fraction of a number of times a first of the pair of alerts is generated relative to a number of times a second of the pair of alerts is triggered based on the relative alert count and the count of the first alert, and displaying the relative fractions in the graphical user interface.

7. A system to manage alerts generated by one or more event sources that run in a distributed computing system, the system comprising: one or more processors; one or more data-storage devices: and machine-readable Instructions stored in the one or more data-storage devices that when executed using the one or more processors controls the system to perform operations comprising: providing a graphical user interface that enables a user to define a dominant alert and select two or more subsumed alerts generated by the one or more event sources; searching log-message files of the one or more event sources for log messages that correspond to the two or more subsumed alerts; generating a single alert message that describes the dominant alert when a log message that corresponds to one of the two or more subsumed alerts is identified; and executing remedial action to correct a problem represented by the dominant alert.

8. The system of claim 7 wherein providing the graphical user interface comprises providing a graphical user interface that enables a user to select a cluster of alerts for the one or more event sources, the cluster of alerts comprises two or more alerts.

9. The system of claim 7 wherein providing the graphical user interface comprises providing a graphical user interface that enables a user to select the dominant alert and one or more subsumed alerts for the event sources, a cluster of two or more alerts comprises the dominant alert and the one or more subsumed alerts.

10. The system of claim 7 wherein searching the log-message files of the one or more event sources for log messages that correspond to the two or more subsumed alerts: identifying log messages of one or more log message files of the one or more event sources with time stamps in a selected time interval; determining the non-parametric tokens of each log message in the selected time interval; and comparing non-parametric tokens of each log message with keywords of each of the two or more subsumed alerts to detect log messages that correspond to the two or more subsumed alerts.

11. The system of claim 7 wherein searching the log-message files of the one or more event sources for log messages that correspond to the two or more subsumed alerts: identifying log messages of one or more log-message files of the one or more event sources with time stamps in a selected time interval; determining the non-parametric tokens of each log message in the selected time interval; comparing non-parametric tokens of each log message with keywords of each of the two or more subsumed alerts to detect a log message that correspond to a dominant alert, the dominant alert is one of the selected two or more subsumed alerts and is the dominant alert; and comparing non-parametric tokens of each log message with keywords of each of the two or more subsumed alerts to detect one or more log messages that corresponds to one or more subsumed alerts, the subsumed alerts are one or more of the two or more alerts.

12. The system of claim 7 wherein searching the log-message files of the one or more event sources that correspond to the two or more subsumed alerts comprises: identifying log messages of one or more log-message foes of the one or more event sources; determining the non-parametric tokens of each log message; comparing non-parametric tokens with keywords that identify an alert; when a log message contains an alert, incrementing an alert count associated with the alert; and for each pair of alerts co-occurring in a sliding time window applied to the one or more log-message files, computing a relative alert count that represents a count of the number of times a pair of alerts are triggered in time limits of the sliding time window; and for each pair of alerts co-occurring in a sliding time window applied to the one or more log-message files, computing a relative fraction of a number of times a first of the pair of alerts is generated relative to a number of times a second of the pair of alerts is triggered based on the relative alert count and the count of the first alert, and displaying the relative fractions in the graphical user interface.

13. A non-transitory computer-readable medium encoded with machine-readable instructions that implement a method carried out by one or more processors of a computer system to perform operations comprising:

determining each type of alert recorded in log messages of one or more log-message files generated by one or more event sources;

determining a relative fraction of a pair of alerts co-occurring within a sliding time window applied to the one or more log-message files;

for each pair of alerts, identifying a first of the pair of alerts as a dominant alert and second of the pair of alerts as a subsumed alert when the relative fraction associated with the pair of alerts is greater than an alert threshold;

generating a single alert message that describes the dominant alert when a log message that corresponds to the subsumed alert co-occurs with the dominant alert; and

executing remedial action to correct a problem represented by the dominant alert.

14. The method of claim 13 wherein determining each type of alert recorded in log messages of one or more log-message files comprises:

identifying log messages of one or more log-message files of the one or more event sources;

determining the non-parametric tokens of each log message;

comparing non-parametric tokens with keywords that identify the type of alert in each log message.

15. The method of claim 13 wherein determining the relative fraction of each pair of alerts co-occurring with the sliding time window comprises:

when a log message contains an alert, increment an alert count associated with the alert; and

for each pair of alerts co-occurring in a sliding time window applied to the one or more log-message files, computing a relative alert count that represents a count of the number of times a pair of alerts are triggered in time limits of the sliding time window; and

for each pair of alerts co-occurring in a sliding time window applied to the one or more log-message files, computing the relative fraction of the number of times a first of the pair of alerts is triggered relative to a number of times a second of the pair of alerts is triggered within time limits of the sliding time window based on the relative alert count of the pair of alerts and the count of the first alert.