Versions Compared

Old Version 4

changes.mady.by.user Rui Liu

Saved on

compared with

New Version Current

changes.mady.by.user Rui Liu

Saved on

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

  • Installation

    1. On Mac, use brew to install maven, imagemagick, rabbitmq. Install mongo either manually or using brew.
    2. git clone the repo.
    3. In the top dir, do "mvn compile" to compile the code, should not take long. Then do "mvn package -Dmaven.test.skip=true -Dmaven.javadoc.skip=true -q".
    4. Start mongod on localhost.
    5. Start RabbitMQ on localhost.
    6. In the top dir, in one shell, run "bin/PolyglotRestlet.sh" to start Polyglot. In another shell tab, run "bin/SoftwareServerRestlet.sh" to start SoftwareServer. If the script fails, change its content from "2.1.0" to "2.2.0". Then you can use the URLs http://localhost:8184 to check Polyglot, and http://localhost:8182 to check SS.

Below are documentation on Polyglot and Softwareserver, including how they work together. Intended for developers or interested users. Started with what Rui learned as part of work of BD-460 "Software Server not setting correct public IP on checkins".

...

  • User Interface

A user interacts with Polyglot. Polyglot works internally with software servers. By default Polyglot runs on port 8184, Softwareserver on port 8182. A user can use "http://<polyglot_ip>:8184/" to see the available endpoints / URLs, such as "http://<polyglot_ip>:8184/servers" to see the server IP list, or "http://<polyglot_ip>:8184/form" to use a form to submit a conversion request.

  • Internal Working

  • The Java Classes
    1. Software Server:
      1. SoftwareServer.java: handles wrapper scripts;
      2. SoftwareServerRestlet.java: handles the Restlet service;
      3. SoftwareServerRESTUtilities.java: the rabbitMQHandler() method handles all the interaction with RabbitMQ.
    2. Polyglot:
      1. The entry point is PolyglotRestlet.java.
        java -cp polyglot.jar:lib/* -Xmx1g edu.illinois.ncsa.isda.softwareserver.polyglot.PolyglotRestlet
      2. Polyglot.java: abstract class;
      3. PolyglotStewardAMQ.java: handles IOGraph and interaction with RabbitMQ;
      4. PolyglotRestlet.java: handles the Restlet interface.
        process_jobs(): at the end, writes the ".url" file.
  • SS Registration

A Polyglot process goes to RabbitMQ, gets the consumer IPs, connects to these IP's softwareserver at the URL "<softewareserver_ip>:8182/applications". If the URL is accessible and contains valid content, Polyglot adds the IP to its server list.

  • SS checkin

A Softwareserver connects to RabbitMQ, picks up jobs (aka msgs) in the queues, processes them, and sends the results back by accessing Polyglot's endpoint at "<polyglot_ip>:8184/checkin/<jobid>/<result_url>".

  • SS capabilities

SoftwareServer uses SoftwareServer.conf + scripts/*/.aliases.txt to configure which applications it will process. For example, SS on dap-dev is configured to convert only demclip and streamclip:

Code Block
languagebash
titleSS conf and .aliases.txt content
collapsetrue
SoftwareServer.conf:
#BatchScripts=scripts/bat
#ShellScripts=/home/polyglot/scripts/sh
#RScripts=scripts/R
#AHKScripts=scripts/csr-configured
#AppleScripts=scripts/applescript
#SikuliScripts=scripts/sikuli
PythonScripts=scripts/py

$ grep -v ^# */.aliases.txt
ahk/.aliases.txt:A3DReviewer
ahk/.aliases.txt:ImgMgk
ahk/.aliases.txt:IrfanView
py/.aliases.txt:demclip_convert.py
py/.aliases.txt:streamclip_convert.py
R/.aliases.txt:PEcAn
sh/.aliases.txt:avconv-audio
sh/.aliases.txt:avconv-video
  • Polyglot REST APIs
  • POL REST endpoints that POL handles on its own without accessing/redirecting to SSes:

    GET:

    /           Returns a list of supported endpoints.

    /alive      Returns "yes".

    /checkin

    /convert    Returns all supported output formats

    /convert/output_format1   returns all supported input formats that can be converted to output_format1

    /convert/output_format1/file_url1   do the conversion: download file_url1

    /form

    /image

    /inputs

    /inputs/<format1>

    /outputs

    /requests

    /servers

    /software

  • POL REST endpoints that access or redirect to SSes:

    GET:

    /file/<file1>      # If file1 doesn't exist and file1.url exists.

    /servers/<server1_ip>[/...]   # Redirects to server1_ip:8182/software/...

    /software/<sw1>    # Accesses all SS:8182/software until finding one that contains sw1 and redirects to <ss_ip1>:8182/software/<sw1>.

 

    POST:

    /servers/<ip1>[/...]   # Redirects to ip1:8182/software/<...>.

    /software/<sw1>        # Accesses all SS:8182/software until finding one that contains sw1 and redirects to <ss_ip1>:8182/software/<sw1>.

  • Implementation Details

  • Configuration files

Both Polyglot and softwareserver SS are implemented in Java, currently using Restlet. Current configuration files are:

...

Softwareserver job checkin is in SoftwareServerRESTUtilities.java, Polyglot accessing RabbitMQ part is in polyglot/PolyglotStewardAMQ.java.

  • Start-up / Initialization
  • When Polyglot starts, it does:
    1. read the configuration file;
    2. start the PolyglotStewardAMQ thread;
    3. start a thread to update Mongo; call PolyglotRESTUtilities.updateMongo(). By default updates every 2 sec.
    4. start the restlet service.
  • When PolyglotStewardAMQ thread starts, it starts 3 threads:
    1. discoveryAMQ(), every 30 s.
    2. process_jobs(), every 3 s,
    3. heartbeat(), to remove unresponsive SSs. every Heartbeat secs, default to 10 s.
  • Two queryEndpoint() methods.
    In SoftwareServerRESTUtilities.java: returns pure text; in PolyglotStewardAMQ.java, returns json.
  • kgm.utility.Utility source code

    https://isda.ncsa.illinois.edu/svn/isda/trunk/kgm/Utilities/src/kgm/utility/Utility.java

Need to install the programs used in Softwareserver scripts:

      convert, unoconv, daffodil, ffmpeg, 7z and 7za, libreoffice, avconv, xvfb-run, eog, flac, ps2pdf, gthumb, htmldoc, kabeja, rar (requires "multiverse"), unzip, unrar, cabextract, ncdump, pdf2djvu, prince, soundconverter, TeighaFileConverter, txt2html, unrtf, cvlc, ebook-convert

which are in the following Ubuntu packages:

      imagemagick, unoconv, (daffodil/), ffmpeg, p7zip-full, libreoffice*, libav-tools, xvfb, eog, flac, ps2pdf, gthumb, htmldoc, (kabeja?), rar, unrar, cabextract, netcdf-bin, pdf2djvu, (prince?), soundconverter, (TeighaFileConverter?), txt2html, unrtf, vlc-nox, calibre

unzip was already installed. ps2pdf is in ghostscript, already installed.

...

To install those that are available on Ubuntu 14.04, Daffodil and copy DaffodilKabeja, do:

Code Block
languagebash
titleCommands to install packages needed by Polyglot and SS
collapsetrue
sudo vi /etc/apt/sources.

...

list   # to uncomment the 4 multiverse lines.

...


sudo apt-get update

...


sudo apt-get install -y imagemagick unoconv p7zip-full libav-tools xvfb eog flac gthumb htmldoc rar unrar cabextract netcdf-bin pdf2djvu soundconverter txt2html unrtf vlc-nox

...

 calibre

cd /home/polyglot

...


sudo scp -pr dap-dev.ncsa.illinois.edu:/home/polyglot/daffodil .

...


sudo chown -R --reference=/home/polyglot /home/polyglot/daffodil

cd /opt
sudo scp -pr dap-dev.ncsa.illinois.edu:/opt/kabeja-0.4 .
sudo chmod 755 kabeja-0.4/kabeja.sh   # Polyglot's Kabeja_convert.sh uses "./kabeja.sh", so needs to be executable.
sudo chown -R --reference=/home/polyglot /opt/kabeja-0.4  # Kabeja_convert.sh copies the input file in it, so needs the write permission.