[PROXYPER] Personal proxy hands out more work units than client receives

Paul Gentle gentleps at muohio.edu
Wed Dec 27 20:07:05 EST 2000


This is a known bug in the client
(http://n0cgi.distributed.net/bugs/show_bug.cgi?id=1475) it is fixed in
v2.8011.464. Check out the Pre-Release Page
(http://www.distributed.net/download/prerelease.html) for a fixed client. We
should be releasing new builds in the near future for pre-release-->release
v2.8012.465. So keep an eye on the Pre-Release page for those and then the
normal download page as well.

Let us know if you have any problems (http://www.distributed.net/bugs/)

Enjoy
moose

Paul Gentle
moose at distributed.net
.plan @ http://n0cgi.distributed.net/cgi/dnet-finger.cgi?user=moose



> -----Original Message-----
> From: owner-proxyper at lists.distributed.net
> [mailto:owner-proxyper at lists.distributed.net]On Behalf Of Edwin ten
> Brink
> Sent: Wednesday, December 27, 2000 2:20 PM
> To: proxyper at lists.distributed.net
> Subject: [PROXYPER] Personal proxy hands out more work units than client
> receives
>
>
> I'm not sure whether this is a personal proxy or a client problem,
> but I discovered an inconsistency when
> examining my personal proxy logs.
>
> The problem is that my personal proxy hands out more packets than the
> client receives (and consequently
> sends back). Client is set to automatically set buffer size and
> determine when to connect. Apparently it
> determines to have 24 work units in the buffers.
>
> What happens seems to be something like this:
> C: request 24 work units
> P: Send 24 work units in one packet
> C: take first 8 work units and consider as one packet, still need 16
> and request them
> P: Send 16 work units in one packet
> C: take first 8 work units and consider as one packet, still need 8
> and request them
> P: Send 8 work units in one packet
>
> Proxy ends up with handing out 24 + 16 + 8 = 48 work units, client
> ends up with 8 + 8 + 8 = 24 work units
> received.
>
> The problem doesn't occur when the proxy has only packets with 8 or
> less work units in it. (Which occurs
> when a lot of fragmented blocks are present.) The clients with the
> following versions do display this
> behavior:
> dnetc v2.8010-463-CTR-00071213 for Win32 (Windows 4.10)
> dnetc v2.8010-463-CTR-00071321 for Linux ELF i386
>
> At least one older Windows client (my proxy reports it as "client
> ver: 710", I'm not sure what version it
> is) seems not affected.
>
> It seems to me to be very inefficient and fragmenting the keyspace
> unnecessarily. Perhaps this could be
> fixed?
>
> Some relevant snippets from a random set of work units from some days
> ago are attached below.
>
> Regards,
>
> Edwin
>
>
> Only the *relevant* lines of the log are displayed.
>
> Proxy Linux x86 build 319
>
> 2000-12-25 08:46:49,rc564 r=128/128, d=0/50, 1.6 Mkeys/sec, tot=2519
> 2000-12-25 08:47:22,Client: Accepted connection from 192.168.1.2:1053
> 2000-12-25 08:47:22,Client: 192.168.1.2 (client ver: 463) is connecting
> 2000-12-25 08:47:22,Client: 192.168.1.2 <-- rc564 x 24 fcf37bd8:80000000
> 2000-12-25 08:47:22,Client: 192.168.1.2 <-- rc564 x 16 fcf37bda:00000000
> 2000-12-25 08:47:22,Client: 192.168.1.2 <-- rc564 x 8 fcf37bdb:00000000
> 2000-12-25 08:47:26,Client: 192.168.1.2 has ended communications.
> 2000-12-25 08:47:26,Client: Closing client connection with 192.168.1.2
> 2000-12-25 08:47:50,rc564 r=80/128, d=24/50, 1.7 Mkeys/sec, tot=2543
>
> Ready 128 -> 80, so 48 sent to client (24 + 16 + 8).
>
> 2000-12-25 12:15:13,rc564 r=152/128, d=0/50, 1.7 Mkeys/sec, tot=2591
> 2000-12-25 12:15:16,Client: Accepted connection from 192.168.1.2:1054
> 2000-12-25 12:15:16,Client: 192.168.1.2 (client ver: 463) is connecting
> 2000-12-25 12:15:19,Client: 192.168.1.2 --> rc564 x 8 fcf37bda:00000000
> 2000-12-25 12:15:20,Client: 192.168.1.2 --> rc564 x 8 fcf37bd8:80000000
> 2000-12-25 12:15:22,Client: 192.168.1.2 has ended communications.
> 2000-12-25 12:15:22,Client: Closing client connection with 192.168.1.2
> 2000-12-25 12:16:14,rc564 r=104/128, d=24/50, 1.8 Mkeys/sec, tot=2615
>
> 2000-12-25 14:29:06,server: --> uplink rc564 x 8 fcf37bda:00000000
> 2000-12-25 14:29:06,server: --> uplink rc564 x 8 fcf37bd8:80000000
> 2000-12-25 14:29:06,server: <-- ack OK rc564 x 8 fcf37bda:00000000
> 2000-12-25 14:29:06,server: <-- ack OK rc564 x 8 fcf37bd8:80000000
>
> 2000-12-25 16:00:32,Client: 192.168.1.2 --> rc564 x 8 fcf37bdb:00000000
> 2000-12-25 16:00:33,server: --> uplink rc564 x 8 fcf37bdb:00000000
> 2000-12-25 16:00:33,server: <-- ack OK rc564 x 8 fcf37bdb:00000000
>
> 24 come back, and are sent to (and acknowledged by) the full server.
>
> ----
>
> Client dnetc v2.8010-463-CTR-00071321 for Linux ELF i386
>
> [Dec 25 08:49:17 UTC] Retrieved 3 RC5 packets (24 work units) from server
> [Dec 25 09:58:32 UTC] Loaded RC5 8*2^28 packet FCF37BD8:80000000
> [Dec 25 11:07:52 UTC] Loaded RC5 8*2^28 packet FCF37BDA:00000000
> [Dec 25 12:17:09 UTC] Loaded RC5 8*2^28 packet FCF37BDB:00000000
>
> 24 work units, in 3 packets of 8 each.
>
> --
> To unsubscribe, send 'unsubscribe proxyper' to majordomo at lists.distributed.net
>
>

--
To unsubscribe, send 'unsubscribe proxyper' to majordomo at lists.distributed.net



More information about the proxyper mailing list