[tahoe-dev] Question about one-node network

Greg Troxel gdt at ir.bbn.com
Tue Jan 4 13:19:40 UTC 2011


Michael Coppola <coppola.mi at husky.neu.edu> writes:

> I currently have a Tahoe-LAFS network that only has one server.  This
> one server is both a storage node and the introducer.  The
> shares.needed, shares.total, and shares.happy values are all set to 1. 
> I plan to add more storage nodes to this network in the near future, but
> I was wondering what would happen to the files I've originally
> uploaded.  Once I add the additional nodes and raise the N, K, and H
> values, will I still be able to access these original files (since the
> files will not have been distributed across the new value of H nodes
> when they were originally uploaded)?  If not, is there a way to tell
> Tahoe-LAFS to replicate the pieces to H nodes after the fact?  Thanks.

You can still get at the files.  Encoding is per file (directory), set at
creation time.  I have files on pubgrid at 3/10 and at 2/7 (reduced to
lack of working servers).

So you'll have directories at 1/1 and immutable files at 1/1.  After you
add 2 more servers (say) and change clients to to 1/3, new files and
directories will get replicated.  But your old directories won't

Then, if your original server fails, you will lose directories and thus
have a hard time unless you've printed out 'tahoe manifest' even for
files on more servers.

Note that H is just a parameter of what's ok to do now, and doesn't
affect the encoding.

To change from 1/1 to 1/5 I think you have to copy and abandon the
original items.

I recommend never using 1/1.  Use 1/5 and then you can repair to spread
to more servers (I have not tested this!).
-------------- 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/20110104/a106f6ac/attachment.asc>


More information about the tahoe-dev mailing list