Suddenly, what are you doing with Java deployment?
Of course, there are many ways to deploy in the end. There is, but in the script language, a light project that can be git cloned with gem, pip, npm, etc. and can solve some dependencies, but when it comes to Java, the fat of the fat jar will surprise you and drop the whole jar file. It will be a little subtle style, such as generating and replacing it manually. (That's not the case! Please let me know if you have a good solution)
In the first place, it seems that you should not write in Java until such a time, but it is said that Java (or JVM language) is lightly required, such as reusing code assets with peripheral tools or wanting to perform parallel computing with a little multithreading. Still exists.
First, let's consider an existing Java deployment method.
It's a way to do something with the same idea as the scripting language. Copy the source code with git or some tool, resolve the dependency with gradle (hereinafter including maven / sbt / etc.) on the execution server and build locally. In some cases, it will be executed as it is with gradle.
It's not a bad method, but it's more like a development server than a deployment destination because it contains the entire set. Also, it's hard to say that it hasn't been modified in the server. I'm worried that if I change and build, and then someone tries to do something and switch to another branch, I don't know.
This is a method to transfer a .class file that is a compilation of .java files to a server and execute it.
Is it a method? I used to do it when I first started using Java. It's easy because you just copy the file built by the IDE. Rather than deploying, it's just a way to run it on a server. If it's a scripting language, it's like copying with .zip or .tar. It is the worst method when considering actual operation, such as not having all .classes, of course not knowing the version well, or mixing.
Collect .class files with IDE or jar command, or transfer .jar files generated by gradle to the server. Transfer the .jar files of the dependent libraries as well.
It's a legitimate method for its primitive nature. In terms of implementation, it is the same as sending a .class file as a zip file, but it is overwhelmingly easy to handle because various Java tools support it.
You can include source code, memos, documents, anything, so I think it's still useful when you can't manage it. However, it's hard to collect all the .jar files that have dependencies. And different versions of .jar are mixed and it becomes Jar hell instead of DLL hell ...
The royal method. Rebuild all dependent libraries as one executable jar. Now that we define main, we can do the following:
> java -jar xxxx.jar arg1 arg2
Recently (?) Seems to be called uber jar or shadow jar.
This is a common technique. The good thing is that you can just transfer a single file. It's all in, so you don't have to worry about it being partially updated and terrible, and you don't have to worry about forgetting the main class.
Personally, it's too fat (fucking big), and when I look only at the production environment, I don't immediately know the name of the library used or the main class I'm running, so I can't say anything unless I unzip it. It doesn't really suit my taste when running on a server. Maybe it's better to operate under proper naming conventions.
It's a bit sad because I sometimes choose fat jar because it's troublesome to collect dependency .jar files such as gradle. It doesn't seem to be that common, but the dependency library is [this way](https://stackoverflow.com/questions/23109276/gradle-task-to-put-jars-from-maven-repository-into-project- It is collected by lib-folder).
You can do the same with fat jars with war and ear files, which you rarely see outside the web. war may be more suitable because it allows finer control, such as including the application configuration file but excluding it from the classpath.
I think the war file is useful if you have many Java EE environments. If it is a war file, it may be acceptable even if the file size is a little large (prejudice)
Create an image by creating an execution environment in the Docker container. How heavy is a Java Docker image? No, Alphine Linux / musc libc / Jigsaw makes it reasonably compact.
Nowadays, you can also create a Docker image with Maven / Gradle at once with Google / Jib.
It will be the favorite from now on. Java's support for Docker is also advancing, and it can be said that the easy-to-understand problems have almost disappeared. It also does version control and you don't have to worry about being tied to Java local behavior.
However, on the other hand, it's Docker ... There are quite a few situations that aren't suitable for Docker, and it's a shame to put a JVM that can control memory relatively straightforwardly in a container.
Looking at various things, we want to solve the following problems for ideal deployment.
What tool meets this requirement? By the way, if you look at other languages, you can usually deploy with gem, pip, npm, etc. The main code can be git cloned or deployed from a private repository. Versions are managed in the repository, and you can see what happened in one shot by looking at the production environment.
Isn't that the case with Java? Given that, at least the repository has a Maven repository. I can't do git clone, which requires compilation, but I feel that support for private repositories is more powerful than other languages. After that, it would be nice if there was a tool. .. ..
That's why I made it.
marun https://github.com/nishemon/marun
A tool to download target artifacts from Maven repositories while resolving dependencies.
> marun install com.example:example:1.0.1
Gradle and maven can resolve dependencies, but they are gorgeous because they are just build tools, and they have the ability to resolve dependencies in dependency libraries (artifacts) and take care of them, but the target There is no feature to download only artifacts. There is a tool called Apach Ivy that is specialized only for resolving dependencies in the Maven repository, but this is also for building in combination with Apache Ant. It is mainly considered to be used for Ivy, and Ivy.xml must be written to use it by itself.
So, I used Apache Ivy to create a tool that downloads only the target artifacts and their dependencies.
I need a private repository, but now, for example, maven-publish plugin and [aws-maven plugin](https://qiita.com/ suzutt / items / bf1e8a8e425a9077b96c) makes it easy to build a Maven repository on Amazon S3.
It's not good to be difficult to use at first, so it is installed with pip. The command line part is Python2 so that it works with system Python as much as possible.
> sudo pip install marun
init Initialize for the time being. Java is required for execution, so please install it first.
> sudo marun init
configuration file is not found!
Your Maven Repository URL []: s3://your_repository
S3 Access Key []:(Enter the access key)
S3 Secret Key []:(Enter the secret key)
Besides s3, you can also use http / https. When you enter it, a configuration file * /etc/marun.conf * will be created.
By the way, Ivy will do his best to download the dependencies of marun itself like this. Due to Amazon S3, there are a lot of jar file dependencies. Dip into * / var / lib / marun *.
Ivy works hard
Download: 130982 / 130982
Download: 1282424 / 1282424
Download: 241622 / 241622
:: loading settings :: url = jar:file:/var/lib/marun/lib/ivy-2.4.0.jar!/org/apache/ivy/core/settings/ivysettings.xml
:: resolving dependencies :: caller#all-caller;working
confs: [runtime]
found jp.cccis.marun#marun;0.1.1 in maven.cccis.jp.s3.amazonaws.com
found com.amazonaws#aws-java-sdk-s3;1.11.475 in bintray/jcenter
[1.11.475] com.amazonaws#aws-java-sdk-s3;1.11.+
found com.amazonaws#aws-java-sdk-kms;1.11.475 in bintray/jcenter
found com.amazonaws#aws-java-sdk-core;1.11.475 in bintray/jcenter
found commons-logging#commons-logging;1.1.3 in bintray/jcenter
found org.apache.httpcomponents#httpclient;4.5.5 in bintray/jcenter
found org.apache.httpcomponents#httpcore;4.4.9 in bintray/jcenter
found commons-codec#commons-codec;1.10 in bintray/jcenter
found software.amazon.ion#ion-java;1.0.2 in bintray/jcenter
found com.fasterxml.jackson.core#jackson-databind;2.6.7.2 in bintray/jcenter
...
commons-logging#commons-logging;1.1.3 from bintray/jcenter in [runtime]
joda-time#joda-time;2.8.1 from bintray/jcenter in [runtime]
jp.cccis.marun#marun;0.1.1 from maven.cccis.jp.s3.amazonaws.com in [runtime]
org.apache.httpcomponents#httpclient;4.5.5 from bintray/jcenter in [runtime]
org.apache.httpcomponents#httpcore;4.4.9 from bintray/jcenter in [runtime]
software.amazon.ion#ion-java;1.0.2 from bintray/jcenter in [runtime]
:: evicted modules:
commons-logging#commons-logging;1.2 by [commons-logging#commons-logging;1.1.3] in [runtime]
---------------------------------------------------------------------
| | modules || artifacts |
| conf | number| search|dwnlded|evicted|| number|dwnlded|
---------------------------------------------------------------------
| runtime | 16 | 2 | 0 | 1 || 17 | 1 |
---------------------------------------------------------------------
It's a good idea to download something from your private repository and run it, but here I'll download Google Closure Compiler.
> sudo marun install com.google.javascript:closure-compiler:+
:: loading settings :: url = jar:file:/var/lib/marun/lib/ivy-2.4.0.jar!/org/apache/ivy/core/settings/ivysettings.xml
:: resolving dependencies :: caller#all-caller;working
confs: [runtime]
found com.google.javascript#closure-compiler;v20181210 in repo1.maven.org
[v20181210] com.google.javascript#closure-compiler;+
found com.google.javascript#closure-compiler-externs;v20181028 in bintray/jcenter
found args4j#args4j;2.0.26 in bintray/jcenter
found com.google.errorprone#error_prone_annotations;2.3.1 in bintray/jcenter
found com.google.guava#guava;25.1-jre in bintray/jcenter
found org.checkerframework#checker-qual;2.0.0 in bintray/jcenter
found com.google.j2objc#j2objc-annotations;1.1 in bintray/jcenter
found org.codehaus.mojo#animal-sniffer-annotations;1.14 in bintray/jcenter
found com.google.protobuf#protobuf-java;3.0.2 in bintray/jcenter
found com.google.code.gson#gson;2.7 in bintray/jcenter
found com.google.code.findbugs#jsr305;3.0.1 in bintray/jcenter
found com.google.jsinterop#jsinterop-annotations;1.0.0 in bintray/jcenter
found com.google.auto.value#auto-value;1.4.1 in bintray/jcenter
found org.apache.ant#ant;1.9.7 in bintray/jcenter
downloading https://repo1.maven.org/maven2/com/google/javascript/closure-compiler/v20181210/closure
...
---------------------------------------------------------------------
| | modules || artifacts |
| conf | number| search|dwnlded|evicted|| number|dwnlded|
---------------------------------------------------------------------
| runtime | 16 | 14 | 14 | 2 || 16 | 16 |
---------------------------------------------------------------------
I feel like this.
> ls -l
total 8
drwxr-xr-x 1 root root 512 December 23 21:37 lib
-rw-r--r--1 root root 4438 December 23 21:37 marun.json
> ls -l lib
total 18980
-rw-r--r--2 root root 3482 February 26 2015 animal-sniffer-annotations-1.14.jar
-rw-r--r--2 root root 2036195 April 12 2016 ant-1.9.7.jar
-rw-r--r--2 root root 74703 November 3 2013 args4j-2.0.26.jar
-rw-r--r--2 root root 1504726 April 7 2017 auto-value-1.4.1.jar
-rw-r--r--2 root root 343222 May 6 2016 checker-qual-2.0.0.jar
-rw-r--r--2 root root 189289 October 31 04:55 closure-compiler-externs-v20181028.jar
-rw-r--r--2 root root 10248761 December 13 03:35 closure-compiler-v20181210.jar
-rw-r--r--2 root root 13162 April 21 2018 error_prone_annotations-2.3.1.jar
-rw-r--r--2 root root 231952 June 15 2016 gson-2.7.jar
-rw-r--r--2 root root 2734339 May 24 2018 guava-25.1-jre.jar
-rw-r--r--2 root root 8764 July 27 2016 j2objc-annotations-1.1.jar
-rw-r--r--2 root root 4075 July 29 2016 jsinterop-annotations-1.0.0.jar
-rw-r--r--2 root root 19943 October 9 2015 jsr305-3.0.1.jar
-rw-r--r--2 root root 1304415 September 7 2016 protobuf-java-3.0.2.jar
You can also run it with marun run
. The Closure Compiler runs, for example, com.google.javascript.jscomp.CommandLineRunner
, but it's a bit clever, so you can do this to find main with the abbreviated name.
> marun run CommandLineRunner
The compiler is waiting for input via stdin.
As you can see in the displayed message, it is almost Apache Ivy. It didn't seem that hard to write on your own if you just wanted to resolve the dependencies, but the deciding factor was that there were pitfalls and that gradle used it in the past. (It seems that they are solving their own dependencies now)
I downloaded the minimum jar file required to start Ivy with Python, and after that I included Ivy with Ivy and tried to resolve the dependency again.
In addition, the contents of marun.json are like this, only a little analysis data is included.
marun.json
{
"1545568648": {
"mains": [
"com.google.javascript.jscomp.CommandLineRunner",
"com.google.javascript.jscomp.LinterMain",
"org.kohsuke.args4j.Starter",
"org.apache.tools.ant.Diagnostics",
"org.apache.tools.ant.taskdefs.KeySubst",
"org.apache.tools.ant.taskdefs.optional.ejb.IPlanetEjbc",
"org.apache.tools.ant.taskdefs.optional.jlink.jlink",
"org.apache.tools.ant.util.ProcessUtil"
],
"dependencies": {
"com.google.code.gson:gson": {
"cache": "/var/lib/marun/ivy/com.google.code.gson/gson/jars/gson-2.7.jar",
"name": "gson-2.7.jar",
"revision": "2.7"
},
...
},
"install": [
"com.google.javascript:closure-compiler:+"
]
},
"context": [
1545568648
]
}
As you can see in the json file, it also helps save and run the library version, but it essentially just downloads all the jars into lib, so you can run it with java -cp lib / *
.
So, I was worried about how to deploy Java, so I tried to make a deployment tool that fills the gap. I think there are some addictive situations, so please use it if you like. Don't expect perfection for now. (I want to do my best)
Also, if you don't have to make such a tool, or if you would like to do this, please let us know!
Recommended Posts