prateek

Install Steps for RMR

These steps have been tested with

• Oracle Linux 6.4
• RHEL 6.5
• CDH 5.1

Note I wish this was parceled up.

Step 0 (all nodes)

prateek

Shortcut for OSX dock to run irssi in tmux with iTerm2

Synopsis

I run irssi inside a tmux session on OSX. I often close the terminal as I usually get notified by growl about important stuff. I don't want to open a terminal and write a command every time I want to check IRC.

The solution

1. Create a shell script with the following content

#!/bin/zsh
/usr/local/bin/tmux attach -d -t irssi || /usr/local/bin/tmux new -s irssi irssi

prateek

---
current_section: Currently
doing_file: ~/Dropbox/notes/wwid-??.md
views:
  times:
    date_format: '%a %_I:%M%P'
    section: All
    count: 0
    wrap_width: 0
    template: '%boldblack%date %boldcyan > %boldwhite%title %boldbgwhite%boldblack%interval%default'

prateek

Videos

• Right from the Wolf's mouth ;) – [1] - Apache Flume @ bigconf.io

[1] http://vimeo.com/83560398

prateek

#!/usr/bin/env ruby
# Uses the Readability Metrics API from http://ipeirotis.appspot.com/readability-api.html
# Accepts text from STDIN (piped) or as an argument
=begin examples
pbpaste|text_score.rb # copy text and run (on OS X) to get the stats for the clipboard.
cat myfile.md|text_score.rb # get scores for the contents of a file
=end

require 'open-uri'
require 'net/http'

prateek

CDH5 Cluster Setup

These are the steps I followed to setup a 6.5 CentOS VM, and install CDH5 and CM5 on it. All these commands should be run on a single node if running on a cluster, it will serve as the master node.

Caveats:

• This is going to use the embedded postgres db for the services, this is a TERRIBLE idea if the enviornment is anything but short lived POC.
• This was done for a 4 node POC cluster where a single instance was going to be the dedicated master - all the CM management and Hadoop master daemons would run on it. And the 3 remaining nodes would be data nodes.

Steps to follow

1. Create new local CentOS 6.5 Image based on this ISO.

prateek

Usage Instructions for Viz-Oozie

    # install graphviz
    $ sudo yum install -y graphviz

    # install vizoozie
    $ git clone https://github.com/iprovalo/vizoozie
    $ cd vizoozie
 $ sudo python setup.py install

prateek

Hadoop Administration Resources

• Official Docs - overwhelming, and invaluable link

• Kathleen Ting: 7 Deadly Hadoop Misconfigurations videoslides

• Philip Zeyliger: The guy who's day job it is to write CM, presented Debugging Distributed Systems - slides

• Gwen Shapira: Scaling ETL with Hadoop slides

• Cloudera's Resource Library link

prateek

Recent versions of Cloudera's Impala added NDV, a "number of distinct values" aggregate function that uses the HyperLogLog algorithm to estimate this number, in parallel, in a fixed amount of space.

This can make a really, really big difference: in a large table I tested this on, which had roughly 100M unique values of mycolumn, using NDV(mycolumn) got me an approximate answer in 27 seconds, whereas the exact answer using count(distinct mycolumn) took ... well, I don't know how long, because I got tired of waiting for it after 45 minutes.

It's fun to note, though, that because of another recent addition to Impala's dialect of SQL, the fnv_hash function, you don't actually need to use NDV; instead, you can build HyperLogLog yourself from mathematical primitives.

HyperLogLog hashes each value it sees, and then assigns them to a bucket based on the low order bits of the hash. It's common to use 1024 buckets, so we can get the bucket by using a bitwise & with 1023:

select

prateek

#!/usr/bin/python
import csv
import sys
import argparse
import io

csv.field_size_limit(sys.maxsize)

parser = argparse.ArgumentParser(description='Clean csv of in-line newlines')
parser.add_argument('infile',help='Path to input CSV file');