BTrace是线上不重启服务的情况下定位问题的神器，本文介绍了最常用的几种场景
定位慢调用
查看com.tim.BtraceCase#count耗时很长时的上下文：入参、成员变量等信息
@BTrace
public class Debug {
@OnMethod(
clazz = "com.tim.BtraceCase",
method = "count",
location = @Location(value = Kind.RETURN)
)
public static void run(@Self Object self, int number, @Duration long duration) {
if (duration / 1000000 > 100) {
println("cost: " + duration / 1000000);
final Object sum =get(field(classOf(self), "sum"), self); //成员变量
println("sum: " + sum);
jstack(); //调用栈
println("----"); //刷新缓存，否则有部分信息没有显示
}
}
}
异常上下文
比如当com.tim.BtraceCase#count方法抛出异常时，捕获异常信息及入参以及对象实例信息。其中关键是location = @Location(value = Kind.ERROR)
@BTrace
public class Debug {
@OnMethod(
clazz = "com.tim.BtraceCase",
method = "count",
location = @Location(value = Kind.ERROR)
)
public static void run(@Self Object self, int number, @TargetInstance Throwable e) {
println("number: " + number); //入参
println(e); //异常信息
final Object sum =get(field(classOf(self), "sum"), self); //成员变量
println("sum: " + sum);
jstack(); //调用栈
println("----"); //刷新缓存，否则有部分信息没有显示
}
}
输出内容：
number: 4100
java.lang.IllegalArgumentException: data should not be divided by 100
sum: 8407050
com.tim.BtraceCase.count(BtraceCase.java:22)
com.tim.BtraceCase.main(BtraceCase.java:10)
----
输出统计信息
http://github.com/btraceio/btrace/blob/master/samples/Histogram.java
