Author: threedayman
Source: Hang Seng LIGHT Cloud Community
Arthas: what is
Arthas is an Alibaba open source Java online diagnostic tool. Arthas supports JDK 6+, Linux/Mac/Windows, command-line interaction, and provides rich Tab auto-completion features to facilitate problem location and diagnosis.
Official website address ** : **arthas.aliyun.com/doc/
Github address: ** : **github.com/alibaba/art…
What can Arthas do
Have you encountered any of the following problems in production or test environments?
- Which JAR is this class loaded from? Why are exceptions related to various classes reported?
- Why is the code I changed not executed? Did I not commit? Wrong branch?
- You can’t debug online when you encounter a problem, can you only add logs and republish?
- There is a problem with the data processing of a user online, but it cannot be debug online or reproduced offline!
- Is there a global view of the health of the system?
- Is there any way to monitor the real-time state of the JVM?
- How to quickly locate the hot spots in the application and generate the flame diagram?
- How do I find an instance of a class directly from within the JVM?
- How to locate slow calling methods.
It says on the official website that Arthas can help solve the above problems, but it does not specify how to deal with them. Here we introduce several commands to explain how to solve the above problems through Arthas.
Quick start to Arthas
Download Arthas
curl -O https://arthas.aliyun.com/arthas-boot.jar
Copy the codeRun Arthas
java -jar arthas-boot.jar
Copy the codeAfter this command is executed, Java processes are displayed. Select the processes that you want to check.
Let’s run the following test code to see Arthas in action
package com.example.arthas;
public class ArthasTest {
public static void main(String[] args) throws InterruptedException {
while(true){
put("test");
}
}
private static String put(String name) throws InterruptedException {
Thread.sleep(1000L);
System.out.println(name);
return "hello "+name;
}
}
Copy the codeSelect the process where our test code runs [3]22756
[INFO] Found existing java process, please choose one and input the serial number of the process, eg : 1. Then hit ENTER.
* [1]: 1200 org.jetbrains.jps.cmdline.Launcher
[2]: 9120
[3]: 22756 com.example.arthas.ArthasTest
Copy the codeWhich JAR is this class loaded from? Why are exceptions related to various classes reported?
Use the sc command to find out which JAR package loaded the class
[arthas@22756]$ sc -d com.example.arthas.ArthasTest
class-info com.example.arthas.ArthasTest
code-source /D:/ideaproject/jsoupservice/target/classes/
name com.example.arthas.ArthasTest
isInterface false
isAnnotation false
isEnum false
isAnonymousClass false
isArray false
isLocalClass false
isMemberClass false
isPrimitive false
isSynthetic false
simple-name ArthasTest
modifier public
annotation
interfaces
super-class +-java.lang.Object
class-loader +-sun.misc.Launcher$AppClassLoader@18b4aac2
+-sun.misc.Launcher$ExtClassLoader@593cfc78
classLoaderHash 18b4aac2
Affect(row-cnt:1) cost in 77 ms.
Copy the codeWhere, the code-source information indicates the class loading path. If it is loaded from the JAR package, the JAR package information will be displayed, as follows
code-source file:/home/trade/u3/u3-mnager/u-mnager-1.jar! The/BOOT - INF/lib/uc - do - 1.0 - the SNAPSHOT. The jar! /Copy the codeThis command can be used to troubleshoot dependency conflicts.
Why is the code I changed not executed? Did I not commit? Wrong branch?
We use the jad command to decompile to see the source information
[arthas@22756]$ jad com.example.arthas.ArthasTest
ClassLoader:
+-sun.misc.Launcher$AppClassLoader@18b4aac2
+-sun.misc.Launcher$ExtClassLoader@593cfc78
Location:
/D:/ideaproject/jsoupservice/target/classes/
/*
* Decompiled with CFR.
*/
package com.example.arthas;
public class ArthasTest {
public static void main(String[] args) throws InterruptedException {
while (true) {
/* 7*/ ArthasTest.put("test");
}
}
private static String put(String name) throws InterruptedException {
/*12*/ Thread.sleep(1000L);
/*13*/ System.out.println(name);
return "hello " + name;
}
}
Affect(row-cnt:1) cost in 164 ms.
Copy the codeVerify that the committed code is running by looking at the decompiled code.
You can’t debug online when you encounter a problem, can you only add logs and republish?
Watch watches the input parameters and return values of the method execution
[arthas@22756]$ watch com.example.arthas.ArthasTest put '{params,returnObj}' -x 2 -n2 Press Q or Ctrl+C to abort. Affect(class count: 1 , method count: 1) cost in 12 ms, listenerId: 3 method=com.example.arthas.ArthasTest.put location=AtExit ts=2021-08-04 17:07:19; [cost=3.10595639883E7ms] result= @arrayList [@object [][@string [test],], @string [hello test], ] method=com.example.arthas.ArthasTest.put location=AtExit ts=2021-08-04 17:07:20; [cost=1009.9733ms] result= @arrayList [@object [][@string [test],], @string [hello test],]Copy the code-x Traversal depth of output result attribute
-n Indicates the number of execution times
Retransform loads the external.class file and retransforms the classes that the JVM has loaded.
Combined with the JAD and MC commands, decompile through the JAD, modify the source code through the editor, and add related logs. MC command to compile the modified code in memory.
Load the new bytecode with the retransform command. Log statements are added by hot loading.
jad --source-only com.example.arthas.ArthasTest > D:/ArthasTest.java
mc d:/ArthasTest.java -d d:
retransform d:/com/example/arthas/ArthasTest.class
Copy the codeThere is a problem with the data processing of a user online, but it cannot be debug online or reproduced offline!
This problem can also be observed by using the above watch method to observe the related call parameters online. Based on the known survey results, you can add the corresponding conditional parameters to the Watch method to filter out the requests that do not need to be observed. Watch command document.
Is there a global view of the health of the system?
Using the dashboard real-time data panel commands, you can view information such as threads, memory, and system parameters. For details about parameters, see the Dashboard documentation
[arthas@17396]$ dashboard -n 1 ID NAME GROUP PRIORITY STATE %CPU DELTA_TIM TIME INTERRUPT DAEMON -1 C1 CompilerThread3 - -1-0.0 0.000 0:0.437 false true -1 C2 CompilerThread2 - -1-0.0 0.000 0:0.375 false true 1 main main 5 TIMED_WA 0.0 0.000 0:0.328 false false -1 C2 CompilerThread0 - -1-0.0 0.000 0:0.312 false true -1 C2 CompilerThread1 - -1-0.0 0.000 0:0.281 false true 23 Arthas -NettyHttptelNetBootstr System 5 RUNNABLE 0.0 0.000 0:0.234 false true 5 Attach Listener System 5 RUNNABLE 0.0 0.000 0:0.031 false True -1 GC task Thread# 7 (ParallelGC) - -1-0.0 0.000 0:0.031 false True-1 GC task thread#6 (ParallelGC) - -1-0.0 0.000 0.031 false true-1 GC task thread#0 (ParallelGC) - -1-0.0 0.031 false true-1 GC task thread#0 (ParallelGC) - -1-0.0 0 0.000 0:0.031 false true-1 GC task Thread #1 (ParallelGC) - -1-0.0 0.000 0:0.031 false true-1 VM Thread - -1-0.0 0.000 0:0.031 false true -1 GC task thread#2 (ParallelGC) - -1-0.00.000 0:0.031 false true -1 GC task thread#3 (ParallelGC) - -1-0.00.000 0:0.031 false true -1 GC task thread#5 (ParallelGC) - -1-0.00.000 0:0.031 false true -1 GC task thread#4 (ParallelGC) - -1-0.00.000 0:0.031 false true 16 arthas-NettyHttpTelnetBootstr system 5 RUNNABLE 0.0 0.000 0:0.015 false true 6 Monitor Ctrl-Break main 5 RUNNABLE 0.0 0.000 0:0.015 false true 2 Reference Handler System 10 WAITING 0.0 0.000 0:0.000 false true 3 Finalizer system 8 WAITING 0.0 0.000 0:0.000 false true 4 Signal Dispatcher system 9 RUNNABLE 0.0 0.000 0:0.000 false true 13 Arthas-timer system 5 WAITING 0.0 0.000 0:0.000 false true 17 Arthas-NettyWebSocketttyBoots System 5 RUNNABLE 0.0 0.000 0:0.000 false true 18 Arthas-NettyWebSocketttyBoots System 5 RUNNABLE 0.00.000 0:0.000 false true Memory Used total Max Usage GC heap 26M 165M 3591M 0.75% gc.ps_scexploiture. Count 2 Exploiture (ms) 16 ps_widow_space 0K 10752K 10752K 0.00% Count 1 ps_old_gen 17M 91M 2693M 0.67% Gc.ps_marksweep. time(ms) 41 nonheap 29M 30M -1 97.08% code_cache 6M 6M 2M 2.57% metaspace 20M 21M -1 97.06% compressed_class_space 2M 2M 1024M 0.25% direct 0K 0K - 0.00% mapped 0K 0K -0.00% Runtime os.name Windows 10 OS. Version 10.0 java.version 1.8.0_291 java.home C:\Program Files\Java\jdk1.8.0_291\jre systemload. Average -1.00 processors 8 timestamp/uptime Fri Aug 06 09:43:41 GMT+08:00 2021/326sCopy the codeClassloader looks at the classloader and the number of classes it loads.
[arthas@17396]$ classloader name numberOfInstances loadedCountTotal BootstrapClassLoader 1 2782 com.taobao.arthas.agent.ArthasClassloader 1 1373 sun.misc.Launcher$ExtClassLoader 1 60 sun.reflect.DelegatingClassLoader 15 15 sun.misc.Launcher$AppClassLoader 1 7 Affect(row-cnt:5) cost in 3 ms.Copy the codeIs there any way to monitor the real-time state of the JVM?
Using JVM commands, you can view the real-time running status of the JVM. It contains machine information, compiler information, garbage collector, class loading information, memory management, thread related information, file descriptor related information. JVM command document
[arthas@17396]$ jvm RUNTIME ----------------------------------------------------------------------------------------------------------------------- Machine-name 17396@WINDOWS-D2B420I jvm-start-time 2021-08-06 09:38:14 management-spec-version 1.2 spec-name Java Virtual Machine Specification spec-vendor Oracle Corporation SPEC-VERSION 1.8 vm-name Java HotSpot(TM) 64-bit Server VM Vm-vendor Oracle Corporation VM-version 25.291-b10 input-arguments-javaAgent :D:\JetBrains\IntelliJ IDEA 2021.1\lib\idea_rt.jar=61057:D:\JetBrains\Int elliJ IDEA 2021.1\ bin-dfile. encoding=UTF-8 class-path C:\Program Files \ Java \ jdk1.8.0 _291 \ jre \ lib \ charsets jar; C: \ Program Files \ Java \ jdk1 8.0 _291 \ jre \ lib \ deploy the jar. C:\Program Files\Java\jdk1.8.0_291\jre\lib\resources.jar; C:\Program Files\Java\jdk1.8.0_291\jre\lib\resources.jar; C: \ Program Files \ Java \ jdk1 8.0 _291 \ jre \ lib \ rt jar; C:\Program ----------------------------------------------------------------------------------------------------------------------- CLASS-LOADING ----------------------------------------------------------------------------------------------------------------------- LOADED-CLASS-COUNT 3914 TOTAL-LOADED-CLASS-COUNT 3914 UNLOADED-CLASS-COUNT 0 IS-VERBOSE false ----------------------------------------------------------------------------------------------------------------------- COMPILATION ----------------------------------------------------------------------------------------------------------------------- NAME HotSpot 64-Bit Tiered Compilers TOTAL-COMPILE-TIME 1317 [time (ms)] ----------------------------------------------------------------------------------------------------------------------- GARBAGE-COLLECTORS ----------------------------------------------------------------------------------------------------------------------- PS Scavenge name : PS Scavenge [count/time (ms)] collectionCount : 2 collectionTime : 16 PS MarkSweep name : PS MarkSweep [count/time (ms)] collectionCount : 1 collectionTime : 41 ----------------------------------------------------------------------------------------------------------------------- MEMORY-MANAGERS ----------------------------------------------------------------------------------------------------------------------- CodeCacheManager Code Cache Metaspace Manager Metaspace Compressed Class Space PS Scavenge PS Eden Space PS Survivor Space PS MarkSweep PS Eden Space PS Survivor Space PS Old Gen ----------------------------------------------------------------------------------------------------------------------- MEMORY ----------------------------------------------------------------------------------------------------------------------- HEAP- memory-usage init: 266338304(254.0 MiB) [MEMORY in bytes] Used: 29895120(28.5 MiB) committed: 173015040(165.0 MiB) Max: 3765960704(3.5gib) no-heap-memory-usage init: 2555904(2.4 MiB) [MEMORY in bytes] Used: 31363048(29.9 MiB) COMMITTED: 32161792(30.7 MiB) Max: -1(-1 B) PENDING-FINALIZE-COUNT 0 ----------------------------------------------------------------------------------------------------------------------- OPERATING-SYSTEM ----------------------------------------------------------------------------------------------------------------------- OS Windows 10 ARCH AMd64 PROCESSORS -count 8 load-average -1.0 VERSION 10.0 ----------------------------------------------------------------------------------------------------------------------- THREAD ----------------------------------------------------------------------------------------------------------------------- COUNT 15 DAEMON-COUNT 14 PEAK-COUNT 16 STARTED-COUNT 21 DEADLOCK-COUNT 0 ----------------------------------------------------------------------------------------------------------------------- FILE-DESCRIPTOR ----------------------------------------------------------------------------------------------------------------------- MAX-FILE-DESCRIPTOR-COUNT -1 OPEN-FILE-DESCRIPTOR-COUNT -1Copy the codeHow to quickly locate the hot spots in the application and generate the flame diagram?
The profiler command supports the generation of flame diagrams of application hotspots. Essentially, you take samples over and over again, and then you collect the results to generate a flame graph (Linux/MAC only).
$ profiler start
Started [cpu] profiling
$ profiler status
[cpu] profiling is running for 4 seconds
$ profiler stop
profiler output file: /tmp/demo/arthas-output/20191125-135546.svg
OK
Copy the codeUse port 3658 by default, arthas, you can open the: http://localhost:3658/arthas-output/ view to arthas – the output directory the profiler results below:
Click to see the detailed results:
How do I find an instance of a class directly from within the JVM?
The VMtool uses the JVMTI interface to query memory objects and force GC.
[arthas@17396]$ vmtool -action getInstances --className java.lang.String --limit 10
@String[][
@String[sun/nio/ch/FileDispatcherImpl],
@String[sun/nio/ch/FileDispatcherImpl],
@String[sun/nio/ch/FileDispatcher],
@String[sun/nio/ch/FileDispatcher],
@String[sun/nio/ch/FileDispatcher],
@String[direct],
@String[sun/nio/ch/FileChannelImpl$Unmapper],
@String[sun/nio/ch/FileChannelImpl$Unmapper],
@String[sun/nio/ch/FileChannelImpl$Unmapper],
@String[sun/nio/ch/FileDispatcherImpl],
]
Copy the codeWith the –limit argument, you can limit the number of returned values and avoid straining the JVM when getting too much data. The default value is 10.
How to locate slow calling methods.
Trace shows the method internal invocation path and prints the elapsed time on each node on the method path
[arthas@18584]$ trace com.example.arthas.ArthasTest bridge Press Q or Ctrl+C to abort. Affect(class count: 1 , method count: 1) cost in 137 ms, listenerId: 1 `---ts=2021-08-06 10:30:11; thread_name=main; id=1; is_daemon=false; priority=5; TCCL = sun. Misc. The Launcher $AppClassLoader @ 18 b4aac2 ` - [1011.4579] ms com. Example. Arthas. ArthasTest: bridge () ` - [1009.9123] ms com. Example. Arthas. ArthasTest: the put () # 13 ` - ts = 2021-08-06 10:30:12; thread_name=main; id=1; is_daemon=false; priority=5; TCCL = sun. Misc. The Launcher $AppClassLoader @ 18 b4aac2 ` - [1002.9228] ms com. Example. Arthas. ArthasTest: bridge () ` - [1002.8016] ms com. Example. Arthas. ArthasTest: the put () # 13Copy the codeThe bridge method is called at 1011.4579ms and 1009.9123ms is called at 1009.9123ms. Therefore, we can analyze the PUT method and trace it if necessary.
Based on the above understanding, use the Arthas command to troubleshoot the previous problems. Arthas provides a variety of command functions. For more information about how to use the command, please refer to the command list
Reference documentation
Arthas user document