[tahoe-dev] [tahoe-lafs] #605: two-hour delay to connect to a grid from Win32, if there are many storage servers unreachable
tahoe-lafs
trac at tahoe-lafs.org
Sun Aug 15 06:16:23 UTC 2010
#605: two-hour delay to connect to a grid from Win32, if there are many storage
servers unreachable
------------------------------+---------------------------------------------
Reporter: zooko | Owner: zooko
Type: defect | Status: assigned
Priority: major | Milestone: eventually
Component: code-network | Version: 1.2.0
Resolution: | Keywords: windows performance hang introduction
Launchpad Bug: |
------------------------------+---------------------------------------------
Comment (by zooko):
#448 is fixed. I looked in the code and this issue #605 is still going to
be unchanged by #448 being fixed because the storage client starts
attempting to connect to each storage server as soon as it learns about
the server: [source:trunk/src/allmydata/storage_client.py at 4131#L102]. One
way to improve the behavior of the client with regard to this ticket would
be to make that call to {{{start_connection()}}} be invoked lazily, only
when the client actually wanted to send a message to that particular
server. However, this would harm alacrity—it would move TCP and foolscap
handshaking from introduction time to request time and thus increase the
delay between making the request and receiving the response (for the first
request made to each server). I don't think that is a good trade-off.
That means this ticket is going to stay open until we implement something
more along the lines of comment:14—a way for clients to get hints that
certain servers are currently off-line without attempting to open a TCP
connection to those servers. Oh, also something that might help with this
would be μTP (#1179).
--
Ticket URL: <http://tahoe-lafs.org/trac/tahoe-lafs/ticket/605#comment:19>
tahoe-lafs <http://tahoe-lafs.org>
secure decentralized storage
More information about the tahoe-dev
mailing list