[tahoe-dev] [tahoe-lafs] #778: "shares of happiness" is the wrong measure; "servers of happiness" is better

tahoe-lafs trac at allmydata.org
Mon Nov 16 13:29:43 PST 2009


#778: "shares of happiness" is the wrong measure; "servers of happiness" is
better
--------------------------------+-------------------------------------------
 Reporter:  zooko               |           Owner:  kevan
     Type:  defect              |          Status:  new  
 Priority:  critical            |       Milestone:  1.6.0
Component:  code-peerselection  |         Version:  1.4.1
 Keywords:  reliability         |   Launchpad_bug:       
--------------------------------+-------------------------------------------

Comment(by kevan):

 I made a tweak to one of the utility functions in upload.py to make sure
 that servers_of_happiness is not overcounted -- if the peer selection
 process yielded both a bucket in self.use_peers and an existing share in
 self.preexisting_shares, and those happened to be different peers, both of
 those peers would be counted. I added a test to demonstrate this behavior.
 My fix was to alter a utility function to check for this situation, and
 only count the peer associated with the bucket if it is detected. If
 anyone has a more elegant fix for this, please post it.

 I also did a bit of cleanup and reorganization. In particular:

   * docs/configuration.txt now uses Zooko's description of this behavior.
   * Various occurrences of 'shares_of_happiness' in comments and docs have
 been updated.
   * test_upload is reorganized; in particular, I updated comments, and
 broke up the problem layouts test a bit.
   * I added a couple of new tests to more thoroughly test things.

 Thoughts?

-- 
Ticket URL: <http://allmydata.org/trac/tahoe/ticket/778#comment:77>
tahoe-lafs <http://allmydata.org>
secure decentralized file storage grid


More information about the tahoe-dev mailing list