Crash and deadlock¶
This page is meant to help a user determine the cause of a program failure and diagnose a program that hangs. In the following, two tools are presented: the Stack Trace Analysis Tool (STAT) and the Abnormal Termination Processing (ATP). Both tools rely on the analysis of the stack backtrace to determine where the application is stalled or view where the application was at the time of the crash.
STAT: The Stack Trace Analysis Tool¶
The Stack Trace Analysis Tool (STAT) is a highly scalable, lightweight tool that gathers and merges stack traces1 from all of the processes of a parallel application. STAT is most effective for diagnosing parallel applications that are hung, i.e. suffers from a deadlock2 or livelock3.
Cray ATP: Abnormal Termination Processing¶
The Abnormal Termination Processing (ATP) is a tool that monitors a running program. In the event of a fatal signal encountered by the program, ATP will handle the signal and perform analysis on the dying application.
Using ATP requires that the target application is built with debug symbols
-g compiler flag).
The next step is to set the
ATP_ENABLED environment variable in you batch
script. It's also recommended to set the maximum size of core files to
A stack trace represents a call stack at a certain point in time, listing the function calls that lead up to the call that caused a problem. ↩
Deadlock is a situation when two threads (or processes) are waiting for each other and the waiting is never ending. ↩
Livelock occurs when two or more processes continually repeat the same interaction in response to changes in the other processes without doing any useful work. ↩