Tuesday, 18 June 2013

One Cap to rule 'em all ...

Wondering which cap it could be?
....Well its Capistrano :D

I am a fan of Capistrano from way back and we use it for almost all kind of deployments - Hadoop, MongoDB clusters and so on.
If you have not tried Capistrano, you must try it and figure out how you can use for deployments in your environment.

Its highly configurable - so capify your stuffs!

Checkout the capified scripts to Deploy a replicated sharded MongoDB cluster on AWS EC2 instances in the following link
Feel free to experiment, report bugs/ issues and contribute back.

For more details, follow the link below :

Monday, 15 April 2013

Monitoring S3 uploads for a real time data

        If you are working on Big Data and its bleeding edge technologies like Hadoop etc., the primary thing you need is a "dataset" to work on. So, this data can be reviews, blogs, news, social media data (Twitter, Facebook etc), domain specific data, research data, forums, groups, feeds, fire hose data etc. Generally, companies reach the data vendors to fetch such kind of data.

        Normally, these data vendors dump the data into a shared server kind of environment. For us to use this data for processing with MapReduce and so forth, we move them to S3 for storage first and processing next. Assume, the data belong to social media such as Twitter or Facebook, then the data can be dumped according to the date format directory. Majority of the cases, its the practice.
Also assuming 140-150GB /day being dumped in a hierarchy like 2013/04/15 ie. yyyy/mm/dd format, stream of data, how do you 
-  upload them to s3 in the same hierarchy to a given bucket?
-  monitor the new incoming files and upload them?
-  save the space effectively on the disk?
-  ensure the reliability of uploads to s3?
-  clean if the logging is enabled to track?
-  re-try the failed uploads?

These were some of the questions, running at the back of my mind, when I wanted to automate the uploads to S3. Also, I wanted 0 human intervention or at-least the least!
So, I came up with 
- s3sync / s3cmd.
- the python Watcher script by Greggory Hernandez, here https://github.com/greggoryhz/Watcher 
A big thanks! This helped me with monitoring part and it works so great!
- few of my own scripts.

What are the ingredients?
  •  Installation of s3sync. I have just used one script of s3cmd here and not s3sync in real. May be in future -- so I have this.
  • Installation of Watcher.
  • My own wrapper scripts.
  • cron
Next, having set up of the environment ready, lets make some common "assumptions".
  • Data being dumped will be at /home/ubuntu/data/ -- from there it could be 2013/04/15 for ex.
  • s3sync is located at /home/ubuntu
  • Watcher repository is at /home/ubuntu
Getting our hands dirty...
  • Goto Watcher and set the directory to be watched for and corresponding action to be undertaken.
  • Create a script called monitor.sh to upload to s3 in s3sync directory as below.
    • The variables you may like to change is s3bucket path in "s3path" in monitor.sh
    • This script will upload the new incoming file detected by the watcher script in the reduced redundancy storage format. (you can remove the header -- provided you are not interested to store in RRS format)
    • The script will call s3cmd ruby script to upload recursively and thus maintains the hierarchy ie. yyyy/mm/dd format with files *.*
    • It will delete the file successfully uploaded to s3 from the local path -- to save the disk space.
    • The script would not delete the directory, as it will be taken care by yet another script re-upload.sh, which acts as a backup for the failed uploads to be uploaded again to s3.
  • Create a script called re-upload.sh which will upload the failed file uploads.
    • This script ensures that the files that are left over from monitor.sh (failed uploads -- this chance is very less. May be 2-4 files/day. -- due to various reasons.), will be uploaded to s3 again with the same hierarchy in RRS format.
    •  Post successful upload, deletes the file and hence the directory if empty.
  • Now, more dirtiest work -- Logging and cleaning logs.
    • All the "echo" created in monitor.sh can be found in ~/.watcher/watcher.log when the watcher.py is running.
    • This log helps us initially and may be later too, to backtrack errors or so.
    • Call of duty - Janitor for cleaning logs. To do this, we can use cron to run a script at sometime. I was interested to run - Every Saturday at 8.00 AM
    • Create a script to clean log as "clean_log.sh" in /home/ubuntu/s3sync
  • Time for cron
    • All set! logging clean happens every Saturday 8.00 AM and re-upload script runs for the previous day, to check if files exist and does the cleaning accordingly.
  • Let's start the script
So, this assures successful uploads  to S3. 
My bash-fu with truth! ;)
Happy Learning! :)

Tuesday, 19 March 2013

Phabricator - pretty()

Ah pretty() -- liking it too much! These days I'm into MongoDB so.. :P
Its like 3am ish now and I know, I'm blabbering ... Gotta sleep!
Ah! please bear with rest of my ramblings!

Last month, I was looking into some of the code review tools!
You know - reviewing the review tools! ;)

Well, I experimented with reviewboard, gerrit, barkeep.
Gerrit does not support post-commit review. I was looking for, a way to conduct code review, after the commit is pushed to git.
Reviewboard is cool. Its existing from a longer time. I experimented a lot on the cloud. Well, the post-review process was little painful with command line.
Barkeep installation is kind of different. I did not invest much time. But it looks interesting!

I went with Phabricator, released by Facebook and Open source too.
Wanna contribute ? Fork here https://github.com/facebook/phabricator/  : else go http://phabricator.org/
Okay I don't wanna geeko-Fi more with operators ;)

Well, Phabricator is used by many companies like Quora (unsure if they are presently using it or not, have a doubt , as I read in some quora thread), DropBox, Path (Android app), Disqus (online commenting system), deviantArt( Painting app -- famous in Chrome web store) .... Ah well, lots more!!

  • What's cool about Phabricator?

It has complete software tools. Ok!! You want ...

    • Code Review ? Differential
    • Post Commit Review ? Audits
    • Code Browser ? Diffusion
    • Bug Tracking ? Maniphest
    • Wiki ? Phriction
    • Want more ? It's under development... So you gonna get more too!
Well, personally, I prefer using Audits, Diffusion, Maniphest. 
Wiki - is used in Confluence/BitBucket -- so for now not essential.
Code Review -- Pre-commit reveiw process stalls the development in agile kinda environment. So not going with it.

  • How's the installation?
I read a question in Quora that why is it difficult to install Phabricator. I believe its not so difficult too. 
May be I will post -- how to install/configure everything with Phabricator. 
Please bear with the mentions about Quora, I'm a quor(a)ddict! :D

  • How to start with Audits?
To get started with the workflow, it does take time, probably because its still under development and documentation demands update. But the greatest thing, I love about Phabricator is, its still under the development and yet, functions -- as it purports to be!! That's Amazing! The UI is lovely too!

I faced this issue -- If you have huge repository, where codereview has never happened before, how do you start the process with previous commits? 
Unfortunately, you cannot do with existing commits! The commits that occur after installation/configuration of Phabricator with that repository, audits are possible.
You will have to create some rules that trigger post commit audits. Its called Herald in Phabricator.

I found a workaround for the old commits -- if you wanna review -- You can browse the repository with Diffusion, goto the module / code you are interested in and click on the commit number. 
Upon clicking on commit number, you have the freedom to post in-line comments/review comments, where even the diff is visible. 
After finishing your commenting process, you can cook it with Raise Concern. I also prefer mapping this with a task tracker -- I mean bug tracker ( Maniphest ) -- add the assignee, link to the comments you gave in description and assign. 
This will send a mail to the team.

Email configuration is one big task. 
I used Postfix with SMTP as configuring outbound mail and have updated in github Phabricator.

Mapping Your Projects/Audits/Team
This takes a lot of time and effort too. 
  • First, you need to create accounts for all. 
  • You will have to configure your remote repository (using git). 
  • Map the modules with the module owners and team lead.
  • Create herald rule which triggers audits.

So, you can set up completely in 2days. But its fun!! Audits can be real fun..!! :)
I will really recommend Phabricator! :D
Happy Tweaking! :)

Wednesday, 20 February 2013

Books that every developer must read!

        I have a rack-space of books! I just wanted to make a wishlist of all the books which I've read and yet to read! :) 

Design Patterns: Elements of Reusable Object-Oriented Software
Erich Gamma Richard Helm Ralph Johnson John Vlissides

Structure And Interpretation Of Computer Programs, Second Edition : By Harold Abelson and Gerald Jay Sussman

Refactoring to Patterns
Joshua Kerievsky

Types and Programming Languages
Benjamin C. Pierce

Code: The Hidden Language of Computer Hardware and Software
Charles Petzold

Object-Oriented Analysis and Design with Applications (2nd Edition)
Grady Booch

Code Complete: A Practical Handbook of Software Construction, Second Edition
Steve McConnell

The Design of the UNIX Operating System [Prentice-Hall Software Series]
Maurice J. Bach

The Pragmatic Programmer: From Journeyman to Master
Andrew Hunt  David Thomas

Practical API Design: Confessions of a Java Framework Architect
Jaroslav Tulach

The Practice of Programming (Addison-Wesley Professional Computing Series)
Brian W. Kernighan  Rob Pike

Programming Pearls (2nd Edition)
Jon Bentley

Writing Secure Code, Second Edition
Michael Howard David LeBlanc

The Mythical Man-Month: Essays on Software Engineering, Anniversary Edition (2nd Edition)
Frederick P. Brooks Jr.

Patterns of Enterprise Application Architecture
Martin Fowler

Introduction to Functional Programming (Prentice Hall International Series in Computing Science)
Richard Bird

The Art of Computer Programming
Donald E. Knuth

Effective Java (2nd Edition)
Joshua Bloch

Thinking in Java (4th Edition) 
Bruce Eckel

Programmers at Work: Interviews With 19 Programmers Who Shaped the Computer Industry (Tempus)
Susan Lammers

Coders at Work: Reflections on the Craft of Programming Peter Seibel

Well, I compiled all the aforementioned from Amazon. I will keep appending to list, if I remember :)
Happy Learning! :)

Monday, 11 February 2013

Hadoop Hangover: How-to launch a hadoop cluster CDH4 [MRv1 / YARN + Ganglia] using Apache Whirr

  This post is about how-to launch a CDH4 MRv1 or CDH4 Yarn cluster on EC2 instances. It's said that you can launch a cluster with the help of Whirr and in a matter of 5 minutes! This is very true if and only if everything works out well! ;) 

Hopefully, this article helps you in that regard.
So, let's row the boat...
  • Download the stable version of Apache Whirr  ie. whirr-0.8.1.tar.gz from the following link whirr-0.8.1.tar.gz
  • Extract from the tarball and generate the key 
  • Generate the key
  • Make a properties file to launch the cluster with that configuration.
  • Now let me tell you how to avoid getting headaches!
    • cluster name: Keep your cluster name simple. Avoid testCluster, testCluster1 etc. ie. No Caps, numerics..
    • Decide on the number of datanodes you want judiciously.
    • Your launch may not be successful, if java is not installed. Make sure the image has Java. However, this properties file takes care of that.
    • It will be good to go ahead with MRv1 for now and later switch to MRv2, when we get a production stable release.
    • This is the minimal set of configurations for launching a Hadoop cluster. But, you can do a lot performance tuning upon this.
    • I had launched this cluster from an ec2 instance, Initially i faced errors, regarding user. Setting the configuration below, solved the problem.
    • Set proper permissions for ~/.ssh and whirr-0.8.1 folder before launching.
  •  Well, we are ready to launch the cluster. Name the properties file as "whirr_cdh.properties".
In the console you can see, links to Namenode and JobTracker Web UI. It also prints how to ssh to the instances in the end.

  • Now, you should be having the files generated. You will be able to see  these files: instances, hadoop-proxy.sh and hadoop-site.xml
  • Starting the proxy
  • Open another terminal, and type
  • You should be able to access the HDFS.
  • You can alternatively download hadoop tarball and launch with 
  •  Okay! So I know that you will not be satisfied unless you a web UI
So, we are good to go! 
  •   If you want to launch MRv2,  use this.
and the same process! 
Happy Learning! :)