I am on a Linux machine and use openjdk7. After finding my code was executed twice as fast when using the -server option, i dove deeper into what was happening inside the VM and found that the Server VM inlines my code like crazy, while the Client VM does not inline at all.
Is this normal behavior?
It is normal behaviour.
The server JVM optimises the code more heavily. This uses more CPU on startup and more memory when it is running.
The client VM is designed to quick start up e.g. applets. It is the default on Windows 32-bit JVMs only.
Related
I am evaluating Jprofiler.
In this process, i tried to attach a profiling scheme to a java process on a remote machine.
it provided a profiling port number, using which could connect to remote process and proceed.
But, once a profiling port is attached, i did not find anyway to de-associate the profiling port from the java process.
I would like to know if there is anyway to remove the profiling port from the java process without killing the java process.
Would appreciate your response.
Thank you very much.
Dev
There is no way to do that because the JVM does not provide a way to unload an agent.
How can I get aheap dump from a remote JVM which runs on linux with WL application server?
When I run locally on a windows machine I know how to get a dump. But, how do I get a dump from the user acceptance test server? Thanks in advance.
You can use JMX to connect to the remote application server (it should be enabled in advance) and use the HotSpotDiagnostic MBean which allows taking a heap dump.
You can use JConsole or VisualVM for invoking the MBean operation.
This post by Mike Haller describes how to use this method with JVisualVM.
Since its a *-nix system, and if you have the necessary privieleges, then it would be easy to connect to using SSH protocol:
Connect to the remote machine:
ssh user#remote-machine-ip-address
Enter the user password once prompted for it (it should be the one for the user on the remote machine and not your current system user).
Generate your heap dump using the jmap utility (JDK binaries path should be availble into your system PATH variable or use a full path to it):
jmap -dump:format=b,file=cheap.bin <pid>
There are three steps:
ssh to your server
ssh <your_user_name>#<remote_ip>
jmap to trigger memory dump
jmap -dump:format=b,file=<your_file_name> <your_jvm_pid>
visualize the heap by jhat (here 512m is the size limit, you can set it depends on the leak's file size, like -J-Xmx2g)
jhat -J-Xmx512m <your_file_name>
jhat -port 7401 <your_file_name>
I write a blog to help analyze performance issue: Performance Optimization
I've used Visual VM successfully for thread dumps and heap dumps, however, you don't list your JAVA version?
JAVA Visual VM is no longer shipped with JAVA, but can still be downloaded here and it's still being maintained. They just did a new minor release: October 19, 2021: VisualVM 2.1.1 Released.
VisualVM has also been distributed in Oracle JDK 6~8 as Java VisualVM.
It has been discontinued in Oracle JDK 9.
Here are steps for connecting to the VM from Dzone, VisualVM: Monitoring Remote JVM Over SSH (JMX Or Not)
For other alternatives, the Baeldung JAVA site, which has great information and tutorials, has A Guide to Java Profilers.
I want to monitor JVM performance on my production environment. I have installed only JRE, not JDK, Hence i can't use jstat, jconsole etc. to monitor the JVM performance.
Can somebody please help to understand how can i monitor JVM performance in this scenario?
Is there any way to achieve this?
(please note that i don't want to monitor it remotely through JMX or something else. i would like to install local agent in each machine which will send the metrics to server at the interval of 1 minute.)
Thanks,
KS
If you manage to get JMX up and running on your VM (from the comment), you can then use jmxterm or jmxfetch to push these JMX metrics into a metrics system (like graphite or Datadog).
If you have enough patience and time to write, you can probably have a look at JVMTI. You can write your code in C/C++ and run it along your Java Process and you can gather information about the JVM without affecting it.
Another simple and naive way is to start your VM with a javaagent written in java but JVMTI is even better than that. The most crucial difference between the javaagent and JVMTI app is derived from the loading mechanics. While the agents are loaded inside the heap, they are governed by the same JVM. Whereas the JVMTI agents are not governed by the JVM rules and are thus not affected by the JVM internals such as the GC or runtime error handling.
You can even give Java Mission Control a try if you're using JDK7 or above :)
Jolokia is a java agent you can use to expose JMX as http. Run jmx2graphite and get those metrics into Graphite. The link includes instructions on Graphite installation (10 minutes)
compared with physical machine, the VM is much more easy to change the machine scale and spec, do you think using VM(VirtualBox, or VMWare) to run benchmark is reasonable or not?
As long as you're only comparing code executed in a VM to other code executed using the same VM program, OS, and virtual architecture, and don't expect the numbers to compare reasonably with benchmarks obtained while running native code in an actual processor, you shouldn't have any trouble using a VM for benchmarking to simulate weaker machines than your actual PC
I want to get a heap dump (suspected memory leak) of a certain Java process. However, when I start the jvisualvm tool, I cannot see any of the running Java processes.
I have Google'd around about this and have already found a couple of articles saying that you have to run the Java processes using the same JDK that you start the jvisualvm tool with in order for it to be able to see them. However, as far as I can see, this is already the case. I'm doing everything locally (I have remote access to the machine).
A couple of things to consider:
The processes are running on a firewalled Windows 2008 server
The processes are running using renamed versions of the JDK java.exe executable
As far as I can see the processes are running using the 1.6.0_18 JDK
One of the running processes starts an RMI registry
I'm waiting on a virtualized copy of the server so I can mess around with it (this is a production server). But in the meanwhile; any ideas as to why I cannot see any of the processes in jvisualvm (or jconsole for that matter)?
Well after I did a little research, it would appear that Peter's comment was correct. Because the JVM processes were launched by another user (the NETWORK SERVICE account because they were being started by a Windows service) they didn't show up in jvisualvm.
Workaround
Since I have access to the application configuration, I have found the following workaround, which involves explicitly enabling unsecured JMX for the target JVM:
Add the following JVM parameters:
-Dcom.sun.management.jmxremote.port=3333 -Dcom.sun.management.jmxremote.ssl=false -Dcom.sun.management.jmxremote.authenticate=false
Add the remote process to jvisualvm using JMX by click File -> Add JMX Connection. You can connect to the process using port 3333. Obviously you can change the port if you want.
Link to article explaining this in a little more detail: http://download.oracle.com/javase/6/docs/technotes/guides/visualvm/jmx_connections.html
Notes
It's probably not a good idea to keep the JVM settings permanently, as they would allow anyone to connect to the JVM via JMX.
You can also add authentication to the JMX JVM parameters if you want to.
The simplest way is to execute jvisualvm as administrator (win: "run as administrator"). Which is not ideal but works. All java processes are visible then.