[tahoe-lafs-trac-stream] [tahoe-lafs] #1382: immutable peer selection refactoring and enhancements
tahoe-lafs
trac at tahoe-lafs.org
Tue Jun 25 20:40:27 UTC 2013
#1382: immutable peer selection refactoring and enhancements
------------------------------------+----------------------------------
Reporter: kevan | Owner: markberger
Type: enhancement | Status: new
Priority: major | Milestone: eventually
Component: code-peerselection | Version: 1.8.2
Resolution: | Keywords: design-review-needed
Launchpad Bug: |
------------------------------------+----------------------------------
Comment (by markberger):
I'm currently figuring out the last few tests that fail with the new
uploader. One of them is test_good_share_hosts in
src/allmydata/test/test_checker.py. Here is a summary of the test:
{{{
N = 4
k = 3
happy = 1
Initial sharemap:
share -> [servers]
0:[A] 1:[A] 2:[A] 3:[A,B,C,D,E]
4 good shares, but 5 good hosts
After deleting all instances of share #3 and repairing:
0:[A,B], 1:[A,C], 2:[A,D], 3:[E]
Still 4 good shares and 5 good hosts
}}}
However, with upload of happiness, repair produces
{{{
0:[A] 1:[A, B] 2:[C, A] 3:[E]
}}}
Is this behavior acceptable? The new algorithm views any additional share
placements over N as redundant and does not attempt to place any more
shares. However, this appears to be different from the old algorithm,
which looks like it tried to place shares on as many servers as possible,
regardless of N (at least from this test).
--
Ticket URL: <https://tahoe-lafs.org/trac/tahoe-lafs/ticket/1382#comment:22>
tahoe-lafs <https://tahoe-lafs.org>
secure decentralized storage
More information about the tahoe-lafs-trac-stream
mailing list