m****r
发帖数: 6639 | 1 我用 jconsole pid, 得到 pid is not a managed VM.
我直接进入jconsole,也找不到任何process.
我的file system是ntfs.
jconsole和我想manage的process都是owned by me.
是不是我需要加什么option,在start process的时候?
谢谢. |
|
B*********h
发帖数: 800 | 2 how did you start JConsole? you need to start jconsole with a java process
ID. |
|
x******o
发帖数: 31 | 3 use windows xp and set TMP to c:\temp (the file system is NTFS).
my user name has no under score ('_').
start jconsole and see nothing in local java process tab.
what may be wrong?
Thanks, |
|
z****e
发帖数: 54598 | 4 jmx用来监控app的情况
用jconsole可以直接接上
jconsole就在jdk的文件夹里面
每一个jdk都有jconsole
如果是开发中debug，用ide可以做很多事
认真区分f789的区别
thread dump如果用了thread管理工具，vert.x, tomcat之类的
应该不太可能用得上
然后log很重要，unit test要做好
最后就是system之间的问题了
这个经验就是不要用各种mq，全部用json，会容易很多
你要想办法能够监听每一个msg，vert.x的shell service非常好用
现在系统越来越大，这些问题越来越多滴给到各个后台的猴子手中
光从无到有做几个网站，没啥意义 |
|
c******n
发帖数: 4965 | 5 many java applications are written as JMX MBeans
but EC2 presents many problems for JMX
1) normally you have only port 22 open, changing security group is a hassle
if you are just debugging for one shot.
2) the second port used by JMX is determined dynamically, so you don't know
which port to open
3) JMX is going to figure out the *internal* ip of EC2 and let your JMX
client to connect to RMI on that IP, which can not be reached.
so to solve these
1) use ssh tunnel, + iptables on client
2) http:... 阅读全帖 |
|
c******n
发帖数: 4965 | 6 【 以下文字转载自 Java 讨论区 】
发信人: creation (努力自由泳50m/45sec !), 信区: Java
标 题: trick to use JMX on EC2
发信站: BBS 未名空间站 (Fri Sep 16 01:46:21 2011, 美东)
many java applications are written as JMX MBeans
but EC2 presents many problems for JMX
1) normally you have only port 22 open, changing security group is a hassle
if you are just debugging for one shot.
2) the second port used by JMX is determined dynamically, so you don't know
which port to open
3) JMX is going to figure out the *internal* ip of EC2 and le... 阅读全帖 |
|
z****e
发帖数: 54598 | 7 沙发说了
你还是下intellij idea的community version吧
eclipse好难用啊，idea的community version足够你用了，正好用来搞core java
jee就不用折腾了，理解轮子的原理就好了
但是core java一定要熟悉，我刚在vert.x俱乐部写了如何检测jvm的各种参数
其中jconsole非常顶用，发给俱乐部邀请给你 |
|
g*****y
发帖数: 36 | 8 【 以下文字转载自 Working 讨论区 】
发信人: ghostly (ghostly), 信区: Working
标 题: 1 opening [Performance Engineer] Bay area
发信站: BBS 未名空间站 (Fri Mar 2 15:01:37 2012, 美东)
请把简历发到m**[email protected].另外经验并不限于lead/principal，初中级的
也要。公司在加州San Mateo。
About the Role
SuccessFactors is seeking a Lead/Principle Performance Engineer who will be
responsible for improving Performance and Scalability of various products
that SuccessFactors deliver.
Duties and Responsibilities:
• Partner with PM and Development tea... 阅读全帖 |
|
x******a
发帖数: 11 | 9 Phone screen 1, focus on Linux and network
1. Configuration files in /etc
2. What command do you use to look at the system performance in linux? CPU/
IO/memory etc? --- top, memstat, ifconfig, netstat, free -m
3. sudo vs su--- in very detailed level
4. how to find current running Java process without using ps command --- /
proc
5. how to monitor a process performance --- JMX, jconsole with my experience
6. ip v4 vs ip v6, tcp vs udp
7. Sth about data center network setting, which is not my domai... 阅读全帖 |
|
Q********3
发帖数: 143 | 10 非大牛
jprofiler, jconsole可以
JNI可能造成泄露，强引用无用的对象会造成内存浪费
OOM一般分为两大类
一是heap OOM，
1.用jprofiler看是那些对象占用了大部分内存，然后分析这些地方是否可以优化
2.调整jvm的内存参数Xms,Xmx
二是permgen space OOM
调整jvm参数，permsize 和maxpermsize
监视内存使用情况可以用MBean，很多大公司都有一套系统去监控CPU，内存，硬盘使用
情况 |
|
w**z
发帖数: 8232 | 11 jconsole
take heap dump
take thread dump
IDE |
|
z****e
发帖数: 54598 | 12 一般问题都不在单机上，问题多数出在系统和系统之间
尤其是涉及到网络的时候，这就需要你对每一个msg做监控
比如某一个node往另外一个node发msg，这个msg是否被记录在log中？
这个时候如果是serializable object的话，就很难看出来了，用json容易很多
jmx和jconsole，就盯着资源使用情况
自己可以用jmx做扩展，比如gc一次，你就在log里面塞入一条记录酱紫
然后jvm本身有很多工具，比如jstat可以用
然后ide也有很多工具可以用
然后你自己写的软件，spring偏简单了，如果是ejb的话
ejb contaienr本身有很多ui可以监控
tomcat也有，spring的话，用aop
vert.x的话，用shell service
总之不管怎样，你要想办法对你系统的每一步都能够看到系统在干嘛
然后做详细的log，最后查起来就方便
另外unit test要做好，test很重要，functional的问题一般都是test没覆盖到
non-functional的问题出了，一般都好交代，可以变出各种理由来
但是functional问题出出来，那就影响kp... 阅读全帖 |
|
g******e
发帖数: 3760 | 13 I bet the application you want to manage did not initialize JMX properly. |
|
B*********h
发帖数: 800 | 14 有。如：
object A已经out of scope了，但是之前被放入一个collection或者map里。如果colle
ction和map是static,那么就算enclosing object被GC了,A还不会被GC.
当然Java并没有严格意义上的memory leak.
用SDK里的JConsole就能检测内存使用情况。如果需要更强的功能，要找个好点的memor
y profiler,如optimizeit. |
|
B*********h
发帖数: 800 | 15 do as goodbug said... there is a very simple profiler included in the JDK, t
he jconsole. check out your JDK bin directory. |
|
B*********h
发帖数: 800 | 16 run the application and get the jps id for the JVM.
start a JConsole(part of JDK, google) and watch how the memory grows. If it
keeps growing boundlessly, you have a problem. A more powerful profiler allo
ws you to monitor all object intances by type and trace back to the code whe
re they are generated. |
|
m******t
发帖数: 2416 | 17
Thread dumps have always been a good starting point for the client project I
'm working on. We run both Websphere and JBoss. In the case of Websphere, it
comes with a set of sophisticated analysis tools. Guess what they try to
analyze - thread dumps, and jvm core dumps (if it was _that_ bad).
provide
here.
JConsole? Provided the server isn't really hung as in "hanging dead". |
|
t*******e
发帖数: 684 | 18 Try connecting JConsole with JBoss, there might be existing runtime mbeans
meeting your needs. |
|
e*****t
发帖数: 1005 | 19 for simply usage, just use jconsole. |
|
c*********e
发帖数: 16335 | 20 是不是这些：
-XX:InitiatingHeapOccupancyPercent=n Percentage of the (entire) heap
occupancy to start a concurrent GC cycle. It is used by GCs that trigger a
concurrent GC cycle based on the occupancy of the entire heap, not just one
of the generations (e.g., G1). A value of 0 denotes 'do constant GC cycles'.
The default value is 45.
-XX:MaxHeapFreeRatio=70 Maximum percentage of heap free after GC to
avoid shrinking.
-XX:MinHeapFreeRatio=40 Minimum percentage of heap free after GC to
avoid ... 阅读全帖 |
|
w**z
发帖数: 8232 | 21 jconsole 只能看，改了参数要重启jvm.
one
'. |
|
g*****g
发帖数: 34805 | 22 4000? You mean -Xmx4000M? Also you want to tweak Xms to
avoid runtime allocation.
BTW, unless you are use 64bits JVM, you can't use more than 4G.
Personally we've used single java JVM to accomodate thousands of
concurrent user sessions. I tend to believe it's your own problem.
You can also use JConsole to check your memory usage. |
|
g*****g
发帖数: 34805 | 23 Did you specify -d64? If you are not dummy, you can estimate
your memory usage beforehand, and if it's enough, it should be
enough.
I would use JConsole or VisualVM to check what's using the
memory. It could be a memory leak. |
|
h***i
发帖数: 1970 | 24 你用jconsole或jvisualvm看一下memory usage不就行了.搞不好还有调NewRatio之类的。 |
|
S***s
发帖数: 104 | 25 Ops need tools to monitor server farm.
They need configurable dashboard view etc.
Developer only need simple tool like jconsole to trouble shooting one
instance. |
|
w**z
发帖数: 8232 | 26 这四种方法我都用过, 从易到难:
visual VM
jconsole has plugin
kill -3
jstack |
|
