MobileRoles

There are several roles needed for a functioning Mobile system.

Terms

• RackApp = The main (currently) Merb Ruby application that makes up "WikimediaMobile"'s core code
• MobileArticle = A custom transformed version of a Wikipedia article in a reusable state

List of Roles

• MobileHTTP = A server ready to handle public-facing mobile requests
• MemcachedServer
• Mysql = A log database server... only used by the LogParse script

Basic Outline of Request Cycle

• An HTTP request comes in to MobileHTTP
• The Proxy Server on port 80 passes the request into the/an RackApp process
  • The RackApp checks with MemcacheServer to see if the MobileArticle is cached
  • If it is not found, then the RackApp uses libcurl to scrape a copy from #{lang}.wikipedia.org and transforms it with libxml2 into a cachable MobileArticle. This is then stored in the cache.
  • The MobileArticle is put in a proper layout by the RackApp and is returned to the user
  • The RackApp writes a log entry in JSON format in a file on MobileHTTP
• The Proxy Server sends back the document

Basic Outline of Log Parsing

• A LogParse script is run every 5 minutes
• The script moves the log file to a tmp location (RackApps are still writing to it)
• The script sends a USR1 signal to the RackApp cluster which causes them to write to a new file
• The script parses the JSON-formatted log file and merges the information into a single stat entry
• The stat entry (makes up a row in the db) is written to the Mysql database
• The temporary log file is deleted