I have been playing around with JMX for the last few days and although I don't mind JConsole or the standard JMX web admin tool, it would be nice to create my own web app that somehow queried the registered JMX MBeans and presented a nice, rich-UI-style "dashboard" with charts and graphs; things that the JMX console or JConsole simply do not have.
Is this possible? I've heard of something called a JMX "agent" but not sure if that has anything to do with how JMX can be queried and managed by a dashboard-style app. Thanks in advance!
Jolokia is an agent based approach to JMX which translates JMX native calls to JSON-over-HTTP. It comes with a Javascript library which is perfectly suited for used within the browser. For a sample usage within a web page look at the demo of the Jolokia-Cubism Integration or even better have a look at the blog post from Tomasz Nurkiewicz which gives a very nice example of using Jolokia along with the Highcharts graphing library.
Said all this, there is even more: Ají, Jolokia's fancy sister, has been started as project which will provide an JMX console bundled along with Jolokia clients. I'm still sorting things outs (well, it moves slowly ...), but I think that over the summer more will be there. One can already play around with it and we still look for volunteers (hint ;-) ....
What you are looking to do has probably already been built unless you really want to do it yourself. Have you looked at VisualVM before? I really like it http://visualvm.java.net/mbeans_tab.html/
To write your own application that manages MBeans remotely I think you could start with this tutorial. Seems to be a good starting point.
http://docs.oracle.com/javase/tutorial/jmx/remote/custom.html
One of the comments here is that VisualVM is a profiler. VisualVM isn't a profiler, it's a platform built on top of the NetBeans platform. The platform supports connecting to a VM in flight a number ways including using a JMXConnector to the MBeanPlatformServer, Serviceability Agent (SA), Attach on demand and so on.
VisualVM does contain 2 different profilers but there is also an MBean plugin that allows you to view and manipulate MBeans that have been registered with the MBeansPlatformServer.
A number of the view in VisualVM scrape data from the MXBeans found in the java.lang.management package that are registered with the MBeansPlatformServer. Many of these MXBeans are wrappers over the SA, a component of HotSpot that maintains a set of performance counters. For example, counters can give you some information about JIT compiler activity, GC activity and even host platform CPU utilization and so on.
You can easily build your own plugins to view or manipulate MBeans. The tutorial mentioned in a posting here is one source of information as to how to get it done. I have written a demonstration VisualVM plugin that looks at data from the MemoryPool MXBeans. That plugin has been augmented by others to do a number of other things. All the code can be found at java.net in the project MemoryPoolView. It covers reading data. However executing functionality is really just executing a method against the JMXConnector with the parameters required by the MBean you are interacting with. IOWs, it's a fairly trivial extension of the plugin.
HawtIO Is about the best plug-and-play, works out of the box solution for getting JMX stuff on a web page. You can put multiple mbeans on the same page and see them all at once, it works in a browser, you don't have to write a bunch of front-end code yourself, and it can be done without opening up weird ports that your IT department will get nervous about. The data visualization options that work with the simple drag and drop interface are pretty minimal, but there is documentation for writing more advanced plugins.
The backend uses Jolokia to put an HTTP REST interface on JMX, which is probably going to be true for any solution you might look at.
Related
I'm on a student research team currently working on a Java project in which we want to be able to gather some info about specific external programs in CentOS 7. Some potential examples of data we might want to gather would be the URL from firefox, or the coordinates from Google Earth. Thus far, looking into how to implement this has mostly led me to using JNA and ptrace.
Is this a good route to take to tackle this problem? And if so, what can I use to find the location of specific information about the programs in memory?
JNA and ptrace are quite low level things. To run system applications you should use:
exec method of Runtime class
Apache Common Exec
Last one is more useful.
But it seems you don't need to run any application at all. To get information from Google Earth you can just use special web API - Google Maps APIs Web Services. If you want to send HTTP request you should try Apache Http Client. It's very powerful library and i'm pretty sure it's all for you need.
PS. More convenient client Java Client for Google Maps Services
PS1. I think java is the worst choice to work with OS and native applications. I see 2 ways to use it:
write agent on C/C++ which will be gather this information for java manager. It's a hard way
use hacks and tricks like here is there a way to get the url from current tab in google-chrome or here Possible to catch URLs in linux?. It's more easy but you will have restricted amount these hacks. And it will work not for all applications and cases.
I have an application that runs within a tomcat (a handful of java servlets) on linux. Now I am tasked to monitor the ressources (such as CPU, memory usage, etc.) within this tomcat (not the single servlets).
I am not a programmer (so I cannot change code) and I am not allowed to change anything within the code. I am also not very familiar with Java (so my apologies if I throw a few things together).
The only thing I can do is start tomcat with java-parameters that activate jmx. So I can remotly check tomcat and the application/servlets running within from remote using JMX.
Question:
What tools (or possibilities) are available to find out what kind of monitoring/ressource information the tomcat (or application with the mbeans) is offering?
I am thinking along the lines of a snmpwalk which would allow me to discover the information provided by snmp. However, I have no idea how to do this with java and jmx (found tons of articles on the internet on how to program it, but that is out of the question).
Thank you very much for your help
You are better off looking at an APM tool to fit this use case. I work for AppDynamics and we do just this. The other common tools are Dynatrace and New Relic. These tools will attach to a Java application at runtime and provide deep monitoring, visibility, transaction tracing, and troubleshooting capabilities. AppDynamics also automatically baselines all of the metrics and alerts you when things start having issues.
I am attempting to create a program that can manage EC2 instances (create, stop, terminate). I am unfamiliar with the service, and after looking through documentation and searching the web have not found any general advice on creating a management service. My questions are:
What programming environment/language would best be suited to creating a management program?
How can this program interact with EC2 instances? (I've looked into AWS EC2 command line tools. Can these be used from a program to create/terminate instances?)
Any general advice in accomplishing this is appreciated (links to examples especially).
What programming environment/language would best be suited to creating a management program?
This question cannot be answered in a definite way, rather you should either choose the language you are most comfortable with or that's best suited to your environment/team instead - there are plenty to choose from, most popular major languages are covered by a dedicated SDK (currently Java, .NET, Node.js, Python, PHP, Ruby), see Tools for Amazon Web Services for the detailed listing and links to further information about each.
Please note that there are also two IDE Toolkits listed that might make it easier to get started if you happen to develop in Java or C#, specifically the AWS Toolkit for Eclipse and the AWS Toolkit for Microsoft Visual Studio.
How can this program interact with EC2 instances?
All the SDKs (and also the command line tools build on top of these) do use the AWS APIs to interact with the respective services (each one has a separate API, but most of them are structured very similar), see Documentation for a listing of all currently available services and links to their documentation:
Welcome to the Amazon Web Services (AWS) Documentation. Whether you are new to AWS or an advanced user, you can find useful information about the services ranging from introductions to advanced features.
To learn how you can get started with AWS, see our Getting Started with AWS guide. If you are interested in learning more about our AWS Free Usage Tier, see our AWS free usage tier article.
I guess you are new to the cloud and aws world, you can use AWS Command line tools to management the services. Also to make things pretty simpler you can make use of the readily available aws management console if you are only worry is to start, stop or terminate the instance. Also people have build some thing called Config management system like opscode chef which is built of ruby or you can use puppet built of puppet lab's custom DSL.
I am not sure if you really need a program to do that. The Amazon Management Console ( https://console.aws.amazon.com ) is pretty straight forward, simple to use, you can spin thousands of instances in a very short time. I created 20 instances in about 3 minutes and deleted(terminated) 9 during my first interaction with the technology. Would you provide a use case where this is necessary? Its like you want to create a missile to kill a chicken(rather than buy a knife)
HTH, Thanks,
Bles
I used to work on a project that deals a lot with EC2 and one of my responsibility was to manage instances remotly( deploy/start/stor/reboot). I used Java and the Amazon API to write a communication module for the application.
In addition I made a few general ant scripts to deal with code that has to be deploed into an instance. Kind a old fashion way, but works perfectly :)
I'm looking for something that can monitor the traffic of a java web application in order to estimate cloud computing prices.
It would be great if it can categorize the traffic in different categories, e.g. database, static resources, pages, etc.
Ideally, it should be something working in the same way New Relic does. Unfortunately, New Relic only monitors times and not traffic...
Does something like this exist?
Thanks.
You don't need any java-specific software for doing that. You can use any network monitor tool and just run your application locally! For monitoring different kind of traffic, I would use different tools too. There are a lot of DB-monitor tools out there... Sorry for not being more specific.
There are quite a lot solutions for monitoring Java web applications. I have tried a few of them but finally stopped at two: Zabbix and JavaMelody. Both are appropriate to monitor and categorize a traffic of apps, although they are completely different in how they work. Zabbix allows watching an app over the long term via JMX. JavaMelody can be built into an app and provides complete insight into business app processes.
Your final decision about Java app monitoring platform depends on the prioritized app features that you’d like to monitor. I recommend you to read the review that helps you look at both solutions in details: http://cases.azoft.com/enterprise-system-monitoring-solutions-business-apps/
We're using Windows 2008 and we are thinking of switching application servers from Adobe ColdFusion 9 to Railo 3.1. This would mean using a new Java servlet container, so instead of Adobe JRun 4, we're looking at Apache Tomcat.
Adobe have a helpful perfmon plugin for CF9. We can gather most stats with that. The problem is, as far as I understand, there is no perfmon plugin for Tomcat.
I wanted to know if there are any kind of free profiling tools we can use to get metrics and performance data on Tomcat, for example requests/sec, memory usage etc.
I don't mind if they are just written to logs so long as we can read them in some format. Also, it doesn't have to be a stand-alone product.
Any and all help appreciated!
Just curious - which application server are you using now? Which one uses perfmon now?
Because you've got to run Tomcat on an operating system - Windows, Linux, etc. You seem to imply that perfmon is useless to you now. I don't believe that's the case.
If you need to embellish info from perfmon, you can certainly buy something. But the cheapest solution for you would be filters that would intercept every incoming request and outgoing response to calculate request counts, response time, etc. You'd write these classes once and declare them in your web.xml. They could write to logs using log4j.
Or maybe Hyperic's solution is what you have in mind. It used to be open source, but Spring bought them a few years back. Then VMWare bought Spring. It's all part of a grander solution.
LambdaProbe will give you monitoring for sessions, memory used, web app sessions and servlets, connections etc.
Take a look at the demo site http://demo.lambdaprobe.org/ for more.
Site login: demo/demo