[tahoe-dev] Object Health

Greg Troxel gdt at ir.bbn.com
Mon Jul 9 22:37:47 UTC 2012


  20 servers total, 17 up consistently.  This is a public grid
  (Volunteer Grid 2), so I don't own most of the servers.

One of the major remaining bugs in tahoe is that repair of a mutable
object increases the version number, combined with old shares being a
cause for repair.  so you have

  18 shares of version 2 placed
  on repair, only 15 are available
  place 18 shares of version 3, and cancel the 15 shares of 2

  [pause, during which 3 servers come back and 3 go away]

  on repair, 15 shares of 3 and 3 shares of 2 are found
  do repair again, making 4

  [pause, during which the 3 come back]

  on repair, 18 shares of 4 are available, and 3 of 3
  do reapir, making version 5



My view is that repair should

  place additional shares of the same version number

  consider old shares as something to be cleaned up, but not a reason to
  increment the version number

With those two changes, you'll get extra shares, but the system should
become stable.

I do see the kind of thing you are reporting with pubgrid, but vg2
should be more stable, and if not the rest of the people should be
seeing the same issues.
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://tahoe-lafs.org/pipermail/tahoe-dev/attachments/20120709/307c3522/attachment-0001.pgp>


More information about the tahoe-dev mailing list