We Keep Coding

Automated updates for Java applications [duplicate]

I would like to implement a java application (server application) that can download a new version (.jar file) from a given url, and then update itself at runtime.
What is the best way to do this and is it possible?
I guess that the application can download a new .jar file and start it. But how should I do the handover, e.g. know when the new application is started and then exit. Or is there a better way to do this?

The basic structure of a solution is as follows:
There is a main loop responsible for repeatedly loading the latest version of the app (if required) and launching it.
The application does its thing, but periodically checks the download URL. If it detects a new version it exits back to the launcher.
There are a number of ways you could implement this. For example:
The launcher could be a wrapper script or binary application that starts a new JVM to run the application from a JAR file that gets replaced.
The launcher could be a Java application that creates a classloader for the new JAR, loads an entrypoint class and calls some method on it. If you do it this way, you have to watch for classloader storage leaks, but that's not difficult. (You just need to make sure that no objects with classes loaded from the JAR are reachable after you relaunch.)
The advantages of the external wrapper approach are:
you only need one JAR,
you can replace the entire Java app,
any secondary threads created by the app, etc will go away without special shutdown logic, and
you can also deal with recovery from application crashes, etc.
The second approach requires two JARs, but has the following advantages:
the solution is pure Java and portable,
the changeover will be quicker, and
you can more easily retain state across the restart (modulo leakage issues).
The "best" way depends on your specific requirements.
It should also be noted that:
There are security risks with auto-updating. In general, if the server that provides the updates is compromised, or if the mechanisms for providing the updates are susceptible to attack, then auto-updating can lead to a compromise of the client(s).
Pushing a update to a client that cause damage to the client could have legal risks, and risks to your business' reputation.
If you can find a way to avoid reinventing the wheel, that would be good. See the other answers for suggestions.

I am currently developing a JAVA Linux Daemon and also had the need to implement an auto-update mechanism. I wanted to limit my application to one jar file, and came up with a simple solution:
Pack the updater application in the update itself.
Application: When the application detects a newer version it does the following:
Download update (Zipfile)
Extract Application and ApplicationUpdater (all in the zipfile)
Run updater
ApplicationUpdater: When the updater runs it does the following:
Stop the Application (in my case a daemon via init.d)
Copy the downloaded jar file to overwrite current Application
Start the Application
Cleanup.
Hope it helps someone.

I've recently created update4j which is fully compatible with Java 9's module system.
It will seamlessly start the new version without a restart.

This is a known problem and I recommend against reinventing a wheel - don't write your own hack, just use what other people have already done.
Two situations you need to consider:
App needs to be self-updatable and keep running even during update (server app, embedded apps). Go with OSGi: Bundles or Equinox p2.
App is a desktop app and has an installer. There are many installers with update option. Check installers list.

I've written a Java application that can load plugins at runtime and start using them immediately, inspired by a similar mechanism in jEdit. jEdit is open source so you have the option of looking to see how it works.
The solution uses a custom ClassLoader to load files from the jar. Once they're loaded you can invoke some method from the new jar that will act as its main method. Then the tricky part is making sure you get rid of all references to the old code so that it can be garbage collected. I'm not quite an expert on that part, I've made it work but it wasn't easy.

First way: use tomcat and it's deploy facilities.
Second way: to split application on two parts (functional and update) and let update part replace function part.
Third way: In your server appliction just download new version, then old version releases bound port, then old version runs new version (starts process), then old version sends a request on application port to the new version to delete old version, old version terminates and new version deletes old version. Like this:

This isn't necessarily the best way, but it might work for you.
You can write a bootstrap application (ala the World of Warcraft launcher, if you've played WoW). That bootstrap is responsible for checking for updates.
If an update is available, it will offer it to the user, handle the download, installation, etc.
If the application is up to date, it will allow the user to launch the application
Optionally, you can allow the user to launch the application, even if it isn't up to date
This way you don't have to worry about forcing an exit of your application.
If your application is web based, and if it is important that they have an up to date client, then you can also do version checks while the application runs. You can do them at intervals, while performing normal communication with the server (some or all calls), or both.
For a product I recently worked on, we did version checks upon launch (without a boot strapper app, but before the main window appeared), and during calls to the server. When the client was out of date, we relied on the user to quit manually, but forbid any action against the server.
Please note that I don't know if Java can invoke UI code before you bring up your main window. We were using C#/WPF.

If you build your application using Equinox plugins, you can use the P2 Provisioning System to get a ready-made solution to this problem. This will require the server to restart itself after an update.

I see a security problem when downloading a new jar (etc.), e.g., a man in the middle attack. You always have to sign your downloadable update.
On JAX2015, Adam Bien told about using JGit for updating the binaries.
Sadly I could not find any tutorials.
Source in German.
Adam Bien created the updater see here
I forked it here with some javaFX frontend. I am also working on an automatic signing.

java 8u31 plugin causes signed applets to load much slower

i have noticed that signed applets are loaded much slower with the latest plugin (included in java 8u31 and 7u75). I have debugged the situation quite a lot and i found out that the problem is directly related to the size of the jar files that are referenced in the jnlp file. The problem is that each time the applet starts, there is some 're-indexing' of the cached jar files that takes time.
To reproduce the issue i did this:
I created a minimal applet and in the jnlp file i used to deploy it, I added several irrelevant .jar files (that are not even referenced, so the classloader does not load them) of considerable size (e.g. 30MB). Of course i am using versioning in the jnlp and capture all http traffic to make sure that the delays are not because of traffic (either re-downloading or certificate revocation checks, etc). I run the applet with the trace enabled and then went through the xml trace log file and found out where the delays come: they are always from the JarSigningVerifier ....
Has anyone else seen something like that?
It is very easy to see and reproduce this behavior and i wonder if there is something i am overlooking. Having worked on applets for the past years extensively, i am totally lost at what can be happening. I can verify that reverting to the previous version of the plugin (and every other version before) works as expected.
I have submitted a bug report with oracle, but i haven't heard back still. Any info or idea will help,
TIA

Same here. I thought already I'm getting crazy. Thanks for sharing this.
We are using Java Web Start, but it's sharing the same problem of re-indexing all jar files (in our case it's an app with quite some jars, so starting takes ages).
Aside from the fact that Oracle suddenly decided to check the certificate of the deploy TLS, which caused some hickup on Linux and Macs (we used a StartSSL certificate there, which isn't included in the Java keystore - on Windows it works as it uses the platforms root certs, too).
And, to make it even worse, on Windows x86 the 8u31 silently dies if -XX:+DoEscapeAnalysis or -XX:+OptimizeStringConcat is present in the JVM arguments, though both parameters are standard in Java 8 (but not in 7, that's why they've been included, still). The 64bit engine doesn't have that problem.
The next thing they changed is they now overwrite the start icons if they've been changed (we changed them to put the 64bit engine's path in there), so it stubbornly changes the path back to the 32bit engine every time.
The behaviour of Oracle is not helpful at all, as they didn't announce any of these changes in their changelogs, let alone announcing the certs changes a few days ahead.
I would like to hear from anyone who's sharing the problems and possible workarounds.
Patric

Have you been able to solve this issue? Have you had a reaction from Oracle?
UPDATE START
I've tried everything I can think of and haven't managed to solve the
issue, so I posted my own question on this issue.
A similar bug report can be found here and is backported to at
least Java 8u51 which I tested. Yet again they managed to increase
startup time for our application.
END UPDATE
We are using Java Web Start for an Eclipse RCP-based application with jars that are all signed.
Startup time is 8s within the IDE, Java version doesn't seem to matter here. With web start the story is different. It becomes (much) worse with every Java update:
7u?? (<60): +2s (10s)
7u60: +5s (13s)
7u75: +15s (23s)
8u31: +12s (20s)
8u40: +21s (29s)
8u51: +23s (31s)

Have you tried your jnlp without versioning? In my experience Java jnlp is very buggy specially if you change the jnlp default values. Versioning support is disabled by support, so try it without versioning and see if it still slow?
For me, I noticed some bugs when I used update="background" value, so I no longer change the default update method. My theory is that Oracle only tests jnlp against default values before they release new Java versions.

BKM to deliver a web-based java application + Jetty + JRE to end-user

I have a Java application with a web-based UI (using Vaadin framework).
I want users to get the app and run on their local machines. Some users can be not very technical, so need to make this as simple and foolproof as possible for them.
Possible solutions:
Create a distributive with the webapp + Jetty + batch and shell scripts to start the app.
maybe include JRE too.
The User Story will be: download some ZIP file, unpack and run "start.bat" or "start.sh". The script will start Jetty and maybe launch a web browser with "http://localhost:8080/myapp" URL. There should be no magic with "installing JRE" or "what's your Java Home" or "Your JRE is too old", etc... This would scare the hell out of many non-technical people.
Difficulties here: what if the machine does not have JRE? I'd like to avoid creating 3 different distributives with JREs (for Win/Linux/MacOS). Packaging all 3 JREs into the same distributive seems like an overkill.
Another approach I see is using Java Web Start. My app would be downloaded through JWS with Jetty JAR as a dependency and start the embedded Jetty in this case.
The User Story: click a button on some website to launch the app locally.
I remember having problems with using JWS from behind a proxy ~5 years ago: the dependencies could not be loaded. Not sure if it's still a problem. I haven't used JWS since then, so I'm very familiar with it.
Another problem with JWS is that it will require people to have internet connection to launch the app (at least for the 1st time or maybe every time), which is not always possible (some organizations can't be connected to the Internet). AFAIK, the last JWS suggests installing JRE if it's not there yet, so this should be easy for end-users.
That's probably it for the potential issues with JWS (?).
Finally, the last approach is to pack everything with something like launch4j or InstallShield or whatever. Hopefully those programs are capable of installing JRE is it's not present yet and also can run in headless mode (for servers).
So far option 2 looks the easiest for most users, assuming that JWS can do what I need.
So, my questions are these:
Which additional pitfalls do you see with these 3 solutions?
Which one would you prefer in my case?

I prefer a combination of the solutions 1 and 3. I mean:
First, build a jar with an embedded Jetty that runs your webapp. This jar must have a main class (that starts an embedded Jetty) and a MANIFEST.MF configured with the "Main-Class" property, so this jar can be runned with: "java -jar myEmbeddedWebApp.jar".
Then, use launch4j to build an exe wrapper for that jar, eg myEmbeddedWebApp.exe

Mirth Running Old JAR File Code

This is really driving me crazy. No matter what I do, it seems that Mirth (1.8.2) is running an older version of my JAR file; I know because of various signs, like:
I can't call any functions
Information logged is not showing up in the logs
Changed log messages are not changed in the log files
Files that were once created and written to in code, but no longer touched by code, are still being created and written
I've tried everything I can think of to make this work. It was working at one point, but now it seems like it's no longer being updated. My process to integrate my changes into Mirth are:
Run an ant script to build the JAR file
Copy the JAR file to \lib\custom
Restart the Mirth service (via Mirth administrator)
I've tried restarting (Mirth service via Services, Java, computer) -- to no avail. I know my JAR file is correct, because I've decompiled it (to make sure it has the latest code) and hashed it (to compare to the hash of the ant-built JAR) -- it is correct and the code is there; it's just not being run.
I'm at wit's end; this occurs infrequently but completely blocks me from developing.
Edit: I also know that my code is correct, because when I run unit tests, it generates the right files and calls the right functions and logs the right information. Only Mirth seems to "not get it."
And my classes are very simple; simple one-argument constructors and a few public methods that return various data. Nothing complex, no nested classes/JARs/dependencies.
Edit: I even deleted my custom JAR file and restarted Mirth, and it's still running my code. Awesome :/ I've added a bounty for this question. I suspect the JAR is cached somewhere (even though they deny it on the Mirth forums) and that cache needs to be cleaned out somehow (although why restarting the Mirth service and my PC doesn't do it is beyond me).
I've also killed all instances of Java (and rebooted my computer), so that makes it highly unlikely that the JVM is caching the JAR somewhere.
I tried reinstalling Mirth. For some reason, it had my custom channel when I booted the administrator for the first time; and infuriatingly, it's still running the old JAR, even though I've updated it with the new one in lib\custom.

I ended up solving this with a combination of actions:
Uninstalling Mirth and Java, and then reinstalling.
I also removed all Java installations except one (one JDK and one JRE).
I stopped Mirth when you copying my JAR. Stop Mirth, copy, and restart; don't try to copy on a live installation. It may or may not pick up your updated JAR depending on if it's loaded into a JVM or not.
This combination of steps seemed to work. Prior to this, I had 3-4 JREs installed (and two JDKs) and I was copying (successfully, according to Windows 7) the JARs while Mirth was running. It's working now!

Mirth is a J2EE application running on a plain old JVM; you have options for debugging it.
You could follow the instructions here for running Mirth Connect via Eclipse. You could then see the JVM classpath, you could set breakpoints and use the debugger.
Mirth is based on the Mule ESB. Mule has its own way of class-loading. You could research it.
If Mirth is really using an old version of your JAR, maybe it's got a cached version around somewhere. Or maybe you made some configuration changes you've forgotten about - perhaps you added a new directory for jars. (Not sure how you do that.)
Mule pays attention to an environment variable named MULE_LIB; maybe that's relevant.
It looks like Mirth Connect 1.8 and Mirth Connect 2.0 have different places for jars (lib/custom and custom-lib, respectively). Which version are you using?

Automatically update jar files

I am currently working on desktop software based on java.It's quite a big code base (more than 40 jar files).
I wish to provide an automatic update functionality. The desktop software constantly checks one back end system to see if there are new versions of the jar files available.
The problem now is: How to replace the updated jar files?

If you deploy your application using Java Webstart (JNLP), you get this mechanism almost for free...
From http://mindprod.com/jgloss/javawebstart.html
The key benefit for Java Web Start is automatic update without having to download the entire program every time.

Easiest would be to check for updates on each startup, download the updates and then launch your application. I think this is the way that Java Web Start works (see aioobes answer).
More complex would be to use either the netbeans or eclipse framework for your application. Both are rather complex and you will have to rewrite your application to work with them. This solution supports live updates.
As far as I am aware there is no easy way to update a running application. It is possible to load new versions of a class with a different classloader, but not possible to unload old versions while they are still referenced.

You can make a little server and a launcher which downloads the newest version, replaces the old one, and starts the jar with:
Runtime.getRuntime().exec("java yourjar -jar");
And you terminate the launcher with:
System.exit(1)

You can also serialize down your state (keep it in memory) and then create a new ClassLoader instance pointing to the new .jar files. Then serialize up your state again using this new classloader. You have just changed the underlaying .jars within a executing product.
Please note that you do not need to change the classloader for everything only for the part that is actually using the .jar files. This can be tricky to conclude what parts that are. And you might get nasty linking errors if done wrongly. So..
.. to keep it simple, use WebStart or a preloader that updates the .jars and then starts the main app (basically what WebStart does for you).
A reason for rolling your own, is that you can use your own format for the .jars, encryption, other packing formats etc.

After reading some answers to many auto-update questions, I thought of a solution. This is how I would implement a secure auto-update for a Java/Kotlin jar application.
Assumption: the installer will contain two jars: a launcher and the main application. Any shortcuts created will point to the launcher, but still be the name of the application. The release will contain the main application and the installer.
The launcher is launched first:
First check if an update has already been downloaded as app_name_update.jar
if an update has been downloaded, rename app_name_update.jar to app_name.jar
Start app_name.jar
This part does not have to be in the launcher, but it's preferred as to not slow down the main application: at this point, the launcher should check for an update (e.g. GitHub releases API) and download it to {CWD}/unverified_app_name_update.jar.
Compare the hash of unverified_app_name_update.jar to an online location containing hashes for all published versions. hashes.txt would be an example found in the same github repository. If the software is open-source, GPG signed commits is a must and the launcher should check if the latest update is a verified commit! If its a proprietary application, keep the hashes.txt at a separate URL from the release where your company does not control the infrastructure (e.g. GitHub).
Main app launched:
No need to check for updates unless updates are mandatory, in which case check for update -> if update found, start loading animation "updating" while you can detect that the launcher is still running. Ensure that the launcher has no race condition while loops!

I found ready project to solve automatically updating.
You can update your app, in your cases you can update jars and resources of your desktop app. The idea of the this is next: wrap you app with starter which can control updating and running you app. In details you can find here.