[freenet-support] 5086- Sucess

2004-07-26 Thread John Huttley
I'm having great success with this, toad.
My home node was previously moribund. Now its as full of life as a
spring lamb.


A minor quibble is on the stats reported on a failure.

I don't have an example on the screen, but it goes like..

Attempts were made to contact 3 nodes
1 cleanly rejected
2 ---
47 backed off

So it sure wasn't trying to talk to just 3 nodes!

--Regards

John

(proudly financing freenet each month)

___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]


[freenet-support] 5086 Funny stats

2004-07-26 Thread John Huttley
Hi toad,
here I've the real thing


Attempts were made to contact 8 nodes.

  * 0 were totally unreachable.
  * 8 restarted.
  * 0 cleanly rejected.
  * 39 backed off.

___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]


[freenet-support] Re: new stable

2004-07-26 Thread Wayne McDougall
vinyl1 [EMAIL PROTECTED] writes:

 Every time we download a new release,
 we hope this will be the one and all the content will zip to our machines with
 lightening speed. 

Ok, forgive me if this has been discussed before (and please point me to this
discussion), but I've been wondering what are realistic expectations of speed?

There is a set of nodes, hoepfully well connected, but all with finite
bandwidth connections. There is some data I'd likem to retrieve, located 
(hopefully) at one or more nodes in a set of finite sized datastores.

It seems NG Routing was introduced to help mediate what I'd over-simplify as
people trying to pull out more than they put in.

So this prompts my borader question: we're pooling bandwidth and data storage in
a large, cryptographically securish network. We're expecting a certain
performance from that network. What sort of speed do we expect from that
network? If the collective data stores are too small, and the collective
bandwidths too limited, then won't we see limited transfers? (Forgive me if
this states the obvious). 

Are we not seeing lightening speeds simply because Freenet as a whole is
underresourced. That maybe this is as good as it gets without more nodes
with more resources?

Has anyone done any modelling? Any ideas? Any comments? What is a reasonable
expectation of Freenet performance right now? Any way to tell? Lower or
upper bounds?

I for one am very happy with what Freenet delivers and see it getting
better steadily albeit with occasional missteps. But my expectations may be
unreasonably low.

I request information and shortly thereafter it arrives (or often immediately)
and in what I'm lead to believe is a relative anonymous way. I like it.


___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]


Re: [freenet-support] 5086 Funny stats

2004-07-26 Thread Trevor Smith
John;

39 backed off means 39 other nodes *would* have been contacted; but due to
max queries or recent rejects; we did not contact them; but rather skipped
them and continued looking for a node that we are still permitted to route
to. (call this load balancing / overload avoidance or whatever you like) so
47 nodes were considered; but we only attempted to contact 8

Trevor

- Original Message - 
From: John Huttley [EMAIL PROTECTED]
To: Freenet Supprt [EMAIL PROTECTED]
Sent: Monday, July 26, 2004 2:50 AM
Subject: [freenet-support] 5086 Funny stats


 Hi toad,
 here I've the real thing


 Attempts were made to contact 8 nodes.

   * 0 were totally unreachable.
   * 8 restarted.
   * 0 cleanly rejected.
   * 39 backed off.

 ___
 Support mailing list
 [EMAIL PROTECTED]
 http://news.gmane.org/gmane.network.freenet.support
 Unsubscribe at
http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
 Or mailto:[EMAIL PROTECTED]



___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]


Re: [freenet-support] 5086 Funny stats

2004-07-26 Thread John Huttley
Thanks for explaining it Trevor.
I thought a backoff happened after attempting to contact them.
I must be confusing it with a reject.

Regards

John

On Tue, 2004-07-27 at 02:50, Trevor Smith wrote:
 John;
 
 39 backed off means 39 other nodes *would* have been contacted; but due to
 max queries or recent rejects; we did not contact them; but rather skipped
 them and continued looking for a node that we are still permitted to route
 to. (call this load balancing / overload avoidance or whatever you like) so
 47 nodes were considered; but we only attempted to contact 8
 


___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]


Re: [freenet-support] Re: new stable

2004-07-26 Thread Toad
I don't think so. There are issues with routing (or there were; there
are more but I'm not sure how to approach them), there are issues with
connections, with the balance between load and routing, and so on...
Most of these can eventually be solved.

On Mon, Jul 26, 2004 at 11:33:07AM +, Wayne McDougall wrote:
 vinyl1 [EMAIL PROTECTED] writes:
 
  Every time we download a new release,
  we hope this will be the one and all the content will zip to our machines with
  lightening speed. 
 
 Ok, forgive me if this has been discussed before (and please point me to this
 discussion), but I've been wondering what are realistic expectations of speed?
 
 There is a set of nodes, hoepfully well connected, but all with finite
 bandwidth connections. There is some data I'd likem to retrieve, located 
 (hopefully) at one or more nodes in a set of finite sized datastores.
 
 It seems NG Routing was introduced to help mediate what I'd over-simplify as
 people trying to pull out more than they put in.
 
 So this prompts my borader question: we're pooling bandwidth and data storage in
 a large, cryptographically securish network. We're expecting a certain
 performance from that network. What sort of speed do we expect from that
 network? If the collective data stores are too small, and the collective
 bandwidths too limited, then won't we see limited transfers? (Forgive me if
 this states the obvious). 
 
 Are we not seeing lightening speeds simply because Freenet as a whole is
 underresourced. That maybe this is as good as it gets without more nodes
 with more resources?
 
 Has anyone done any modelling? Any ideas? Any comments? What is a reasonable
 expectation of Freenet performance right now? Any way to tell? Lower or
 upper bounds?
 
 I for one am very happy with what Freenet delivers and see it getting
 better steadily albeit with occasional missteps. But my expectations may be
 unreasonably low.
 
 I request information and shortly thereafter it arrives (or often immediately)
 and in what I'm lead to believe is a relative anonymous way. I like it.
 
 
 ___
 Support mailing list
 [EMAIL PROTECTED]
 http://news.gmane.org/gmane.network.freenet.support
 Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
 Or mailto:[EMAIL PROTECTED]

-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.


signature.asc
Description: Digital signature
___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]

Re: [freenet-support] Connection bug in recent builds

2004-07-26 Thread Toad
On Sun, Jul 25, 2004 at 12:14:41PM +0200, Marc wrote:
 Hi toad, you wrote
  Recent builds, both stable and unstable, have a bug that causes
  connections to fail with crypto related errors. This is probably not due
  to NativeBigInteger, as it happens on stable, which doesn't have NBI.
  Iakin has produced it reliably, although most nodes seem to work anyway.
  It can be reliably reproduced on a local test network. I have made some
  progress (an identity appears to be corrupted somehow), but I haven't
  finished yet. I will fix it on Monday.
 
 Do these errors include the following? This doesn't look crypto related.

It is. The other end got a crypto error.
 
 25.07.2004 11:40:14 (freenet.support.io.NIOInputStream, YThread-2822, NORMAL): 
 waited more than 12ms in NIOIS.read() tcp/connection: 
 8481127.0.0.1:36693,[EMAIL PROTECTED]:[EMAIL PROTECTED] closing
 java.lang.Exception: debug
 at freenet.support.io.NIOInputStream.read(NIOInputStream.java:302)
 at 
 freenet.interfaces.FreenetConnectionRunner.handle(FreenetConnectionRunner.java:81)
 at 
 freenet.interfaces.LocalNIOInterface$ConnectionShell.run(LocalNIOInterface.java:268)
 at freenet.thread.YThreadFactory$YThread.run(YThreadFactory.java:285)
 
 I used ethereal, but forgot to save. So from memory this is what
 happens: Frost sends a ClientHello message, but gets no answer and after
 2 minutes, fred resets the connection.

This is on FCP? Ouch. Is the node/machine overloaded (CPU-wise?)?
-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.


signature.asc
Description: Digital signature
___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]

Re: [freenet-support] start-freenet.sh patch to disable NPTL on SuSE Linux

2004-07-26 Thread Toad
On Sun, Jul 25, 2004 at 03:20:04AM +0200, Nomen Nescio wrote:
 On Sat, Jul 24, 2004 at 11:06:23PM +0100, Toad wrote:
  Will be committed to unstable soon.
 
 Having read http://people.redhat.com/drepper/assumekernel.html I think
 we can do away with the distribution check and set LD_ASSUME_KERNEL to
 2.4.1 for all Linux distributions.  Patch for start-freenet.sh below.

No, we can't. Even 1.5.0-b2 crashes without it.
-- 
Matthew J Toseland - [EMAIL PROTECTED]
Freenet Project Official Codemonkey - http://freenetproject.org/
ICTHUS - Nothing is impossible. Our Boss says so.


signature.asc
Description: Digital signature
___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]

[freenet-support] Weekend releases!?

2004-07-26 Thread Rudolf Krist
Hello Toad!
I think that it is a very bad Idea to publish a new official build, when 
in one day or two you will go to weekend, especially if it is a stable 
build! When there is a serious bug, it will take much longer to solve 
the problem. And the network can rather go broken, like -- I think -- it 
happens now! The longer my node runs, the more RNFs I get. Im using the 
8086 build.

Connections open (Inbound/Outbound/Limit)   54 (13/41/150)
before update(8085/8086) I usually had a value about 140
http://localhost:/servlet/nodestatus/version_data.txt
# Histogram of node versions in fred's Routing table
# 25.07.2004 19:18:20
# nodes: 291
Fred,0.5,STABLE-1.50,5084   135
Fred,0.5,STABLE-1.50,5085   82
Fred,0.5,STABLE-1.50,5086   74
When you think that you are ready to publish a new build, it is 
Thursday, Friday or Saturday and you know that you want to have some 
free days, without development/support, maybe it will by better to wait 
until it is Monday to publish the new stable build?!

Regards
Rudi
--
Rudolf Krist
Mail: rudi.krist -at+ web.de
PGP Key Id: 0x5C1708B8
___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]


[freenet-support] CPU pegging at 100% since 5085

2004-07-26 Thread Mike Z




Since upgrading to 5085 (and 5086), Freenet is using up a huge amount
of processor time, to the point where the web interface no longer even
responds. This doesn't happen every start, but once it occurs, freenet
has to be killed and restarted. It sometimes happens when first
started, sometimes after running for a while.

The machine is running Win XP, with a 3.2Ghz P4 with HT, and 1GB of
RAM. Both processor threads are pegged at full utilization, and freenet
is holding steady at about 150MB of RAM used. 

Any help would be appreciated.

Uptime:  0 days,  1 hour,  51 minutes 
Current routingTime: 0ms. 
Pooled threads running jobs: 272   (54.4%)   
[Rejecting incoming connections and requests!]  
Pooled threads which are idle: 10
It's normal for the node to sometimes reject connections or requests
for a limited period. If you're seeing rejections continuously the node
is overloaded or something is wrong (i.e. a bug).
Current estimated load for rate limiting: 452.2%. 
Load due to thread limit = 54.4%
Load due to routingTime = 10% = 100ms / 1000ms = overloadLow (80%)
Load due to messageSendTimeRequest = 452.2% = 4522ms / 1000ms 
overloadLow (80%)
Load
due to output bandwidth limiting = 14.6% because outputBytes(215153)
= limit (1474560 ) = outLimitCutoff (2) * outputBandwidthLimit
(12288) * 60
Load due to expected inbound transfers: 2.4% because:
404.76961565213924 req/hr * 0.030716723549488085 (pTransfer) * 377327.0
bytes = 4691380 bytes/hr expected from current requests, but
maxInputBytes/minute = 2949120 (output limit assumed smaller than input
capacity) * 60 * 1.1 = 194641920 bytes/hr target
Current estimated load for QueryRejecting: 55.4%.


___
Support mailing list
[EMAIL PROTECTED]
http://news.gmane.org/gmane.network.freenet.support
Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support
Or mailto:[EMAIL PROTECTED]