Tumblr interview:
#Write a function for finding an ordered word pair occuring with the greatest frequency in a given string.  For example
 
# "I want to be a part of it, New York, New York."
 
#"I want", "want to", "to be", "be a",...
 
#And the most frequent is "New York".
 
def frequency
  source="I want to be a part of it, New York, New York."
 
  source = source.gsub(",","")
  source = source.gsub(".","")
  list=source.split(' ')
  counts={};
  prev=""
  list.each do |word|
    if prev !=""
      if counts[ "#{prev} #{word}" ] == NIL
        counts[ "#{prev} #{word}" ] =0
      end
      counts[ "#{prev} #{word}" ] +=1
    end
    prev=word
  end
  print counts
  print counts.sort{|a,b| a[1] <=> b[1]}.last
end
 
frequency
 
1)            Assume the following sequence of bash commands.  Explain what each does, whether it can fail, how it can fail
> and explain the result.  
>
> echo *
> echo "*"
> echo '*'
> echo `*`
>
> mkdir a
> cd a
> touch b
> ls > c
> cat c
 
2) Given some bash commands how do you redirect to make a new file? And will the new file set show up?
 
This will be asked if you fail to answer correctly to final line in 1)
>
> 3)        "given two words write a function that returns true
if the letters in one is a permutation of another".  eg if two words are "abc" and "acd" it should return true.  The answer involves sorting
 
 
> 4)        If you have a master and a number of standby slaves, describe the
process to promote one of the slaves to a master safely allowing for the fact the just-failed master might come back online.
 
 
> 5)        Make sure you understand bash syntax '>' at a system level.
What does it do?  Write a c or perl program to emulate the process of using '>' in bash.
 
 
> 6)        Write a function that will linearly roll out a feature on a website over
24 hour period.  user id 1 should be rolled out last.  You are given the total number of users.  Assume that this function will be used by a web application.  Do not assume that all users will be visiting it during 24 hours period.
 
 
> 7)        If you have a large amount of servers and applications running on the servers, how can you aggregate/ roll out  reshooting of the entire DB scheme? Can you write a function to do that?
 
This problem I did not understand fully.  Perhaps read up on tumblr's
github "jetpants" code.   Something like 'given N application servers
and N aggregators that these application servers are writing distributively (e.g. aggregator 1 is responsible for keys starting with A and aggregator 2 is for B and C) give a strategy to rebalance the keyspace when a new aggregator is added.  You can assume that the app servers have local caches.  Strategy must not use time or puppet.
 
 
 
 
> 8)        What do you think of XYZ technology? What is your opinion etc?
These are the ones I can recall. They were similar to the test questions you sent. Hope they’ll be helpful to the other candidates
 
=========================
Find all anagrams (a) in the list (s) of any given word (c)
 
s={"rodo","rood", "odor", "desk", ...}
c={"door"}
a={"rood", "rodo", "odor”}
 
Solution: Sort characters of each array item in (s) and compare it to the sorted characters in (c)
=========================
 
 
Given a sequential list of user_ids, a start and stop time, and assuming a consistent rollout rate, write a function to determine whether the feature is currently enabled for a given user_id.
 
Solution: Map user_ids to the exact second each user_id is supposed to receive the feature, compare requested user_id with the map
 
=========================
The rest were about my experience with databases, CDNs and server optimization.