RE: Longtime idle connections, was: [freenet-support] Build 659 (unstable)

2003-02-12 Thread Niklas Bergh
And here comes the third dump of the ocm contents

155.239.180.179:34467 0 0 yes 1 0 572922365 QThread-12740 Inbound 33c0b6
80.134.237.85:21321 0 0 yes 2 0 533401428 QThread-25929 Inbound 7e5c44
80.134.237.85:15291 0 0 yes 8 0 484932664 QThread-34344 Inbound 5e833f
68.82.148.48:18230 0 0 yes 1 0 450711357 QThread-36576 Inbound 514847
129.240.200.147:39076 1 147772 yes 471 0 198812426 QThread-45337 Inbound
3c54ad
66.28.14.56:4732 1 1021207 yes 1740 0 173207288 QThread-45442 Inbound
6a5180
210.86.48.242:44180 0 0 yes 1 0 106259012 QThread-45049 Inbound f955a
134.193.75.29:3277 0 0 yes 11 0 198269616 QThread-45215 Inbound 1423d4
216.87.102.230:2816 0 0 yes 8 0 73175861 QThread-45395 Inbound 4e23b2

Please note that the same connections are still there.. plus some
newcomers. The send-queue-size 147772 is also stalled (haven't changed
in a couple of hours)

Summary of current connections:
Number of open connections: 120
Number of outbound connections: 1
Number of inbound connections: 119

Routing time status:
# Current routingTime: 0ms.
# Active pooled jobs: 276 (92.0%)   [QueryRejecting all incoming
requests!] 

Threadcount (maximumthreads is set to 300...):
Total pooled threads 349
Available pooled threads 74
Pooled threads in use 275

And yes.. The node is just about idle now.. 

A couple of strange things I note though..
1. The node is not *completely* idle.. There seems to be one or two
threads that are still making themselves useful, accepting an incoming
connection now and then and establishing a new outgoing connection once
in a while. I assume that given enough time they will also be stuck
however...

2. For some really strange reason: 'Local mean traffic (queries per
hour) : 19399.68744947998'. This CANNOT BE TRUE.. My node is definitely
not serving up any 20kQueries/h, close to no processor or bandwidth is
used. Is the open-but-idle connections counted into this value for some
reason?!

/N






-Original Message-
From: Matthew Toseland [mailto:[EMAIL PROTECTED]] 
Sent: den 11 februari 2003 10:02
To: Niklas Bergh
Subject: Re: Longtime idle connections, was: [freenet-support] Build 659
(unstable)


On Tue, Feb 11, 2003 at 09:52:17AM +0100, Niklas Bergh wrote:
 
 
 -Original Message-
 From: Matthew Toseland [mailto:[EMAIL PROTECTED]]
 Sent: den 10 februari 2003 19:55
 To: Niklas Bergh
 Subject: Re: [freenet-support] Build 659 (unstable)
 
 
 On Mon, Feb 10, 2003 at 11:38:09AM +0100, Niklas Bergh wrote:
  659 seems to have solved the 'Node suddenly ceases traffic'-problem.

  I
 
  have a node that has been up for close 120 hours now and there is 
  till
 
  no sign of stagnation. It is serving up 13000 q/h at about 1MBit/s.
 
 Yay. 1Mbps _both ways_ ?
 
 Nope, summed up. Most of it (80% or so) is outgoing traffic.
 
 
  
  One thing though:
  
  Peer addr Send Count Send Size Receiving Messages Idle time Lifetime
  Thread Type ID 155.239.180.179:34467 0 0 yes 1 0 393188783 
  QThread-12740 Inbound 33c0b6 80.134.237.85:21321 0 0 yes 2 0
353667846
 
  QThread-25929 Inbound 7e5c44 80.134.237.85:15291 0 0 yes 8 0 
  305199082
 
  QThread-34344 Inbound 5e833f 68.82.148.48:18230 0 0 yes 1 0 
  27095
  QThread-36576 Inbound 514847 213.208.127.216:17565 0 0 yes 7 0 
  172335455 QThread-44615 Inbound 134183 64.252.121.1:34381 0 0 yes 10
0
 
  22933808 QThread-45372 Inbound 323c46 216.15.192.235:56589 0 0 yes 2

  0
 
  42376945 QThread-42938 Inbound 13374f 134.193.75.29:3277 0 0 yes 8 0
  18536034 QThread-45215 Inbound 1423d4 129.240.200.147:39076 1 147772

  yes 471 0 19078844 QThread-45337 Inbound 3c54ad
  
  This is the initial entries of my ocm table. I do not like that
  idle-time always is zero on those connections where 'receiving' is 
  'yes'. It seems like the idletime is set to zero even if the 
  connection is idle. Should it really be so, couldn't idle-time be
set 
  to indicate the time since the last packet was recieved on that 
  connection instead?!
 
 Err... hmm. It is transferring a message with trailing fields i.e. a
 file. This can take a long time, but it is not idle during that 
 time...
 
 Is the idle time specified in milliseconds? In that case, over the 
 first connection, one message in 109 hours
 
 OCM 24 hours later:
 
 Peer addr Send Count Send Size Receiving Messages Idle time Lifetime 
 Thread Type ID 155.239.180.179:34467 0 0 yes 1 0 470074608 
 QThread-12740 Inbound 33c0b6 80.134.237.85:21321 0 0 yes 2 0 430553671

 QThread-25929 Inbound 7e5c44 80.134.237.85:15291 0 0 yes 8 0 382084907

 QThread-34344 Inbound 5e833f 68.82.148.48:18230 0 0 yes 1 0 347863600 
 QThread-36576 Inbound 514847 213.208.127.216:17565 0 0 yes 7 0 
 249221280 QThread-44615 Inbound 134183 134.193.75.29:3277 0 0 yes 8 0 
 95421859 QThread-45215 Inbound 1423d4 129.240.200.147:39076 1 147772 
 yes 471 0 95964669 QThread-45337 Inbound 3c54ad
 66.28.14.56:4732 1 1021207 yes 1740 0 70359531 QThread-45442 Inbound
6a5
 
 Please note that this is the same connections as yesterday

Re: Longtime idle connections, was: [freenet-support] Build 659 (unstable)

2003-02-12 Thread Niklas Bergh
I am not totally sure (starting the test today) but I *think* it does. I
remember that I started noticing this problem when I first increased
maxmimumThreads in order to make my node use up more of my available
bandwidth

/N


___
Support mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org:8080/mailman/listinfo/support