In message <20021115020548.GC7038@servalan>, Matthew Toseland
<[EMAIL PROTECTED]> writes
On Sun, Nov 03, 2002 at 06:01:03PM +0000, Roger Hayter wrote:
In message <[EMAIL PROTECTED]>, Greg Wooledge
<[EMAIL PROTECTED]> writes
>John E. Mayorga ([EMAIL PROTECTED]) wrote:
>
>>I have downloaded a
>>seednodes.ref file from
>>http://wooledge.org/~greg/seednodes.ref to no avail (I
>>believe he is using version .4).
>
>I'm running CVS branch rel-0-5-1, which I compile every day or so,
>in the hopes that it will one day magically work great. I take
>"snapshots" of my seednodes.ref every once in a while when my node
>appears to be having a good run (e.g., my routing table hasn't shrunk
>to 5 nodes).
>
>> 4 were totally unreachable.
>>0 restarted.
>>1 cleanly rejected.
>
>Have you checked your node reference status page yet?
>(:8888/servlet/nodestatus/nodestatus.html) It might give some
>useful details.
>
>> 0 were totally unreachable.
>>0 restarted.
>>0 cleanly rejected.
>
>I (sarcastically) refer to that as my Favorite Error Message of All
>Time. That's a very bad sign, but unfortunately I don't know any
>cure.
>
>>ipAddress=24.127.54.220
>
>>listenPort=55555
>
>I'm able to establish a connection to that IP/port from here. That's
>good.
>
>>%transient=false
>
>Good.
>
>I can't think of anything else to tell you than "be patient, keep
>retrying, and if your routing table shrinks below 10 nodes, either
>restart Freenet or try someone else's seed nodes".
>
AMI, my node was down to 1 reference before I went out for the PM, now
it is back to 17, 8 of them contacted. This is with build 525 (which we
are asked not to use because of some routeing problem); if I use a
later build it just leaks threads until it dies. In this situation, new
What do you mean? The cpu overloading problem was fixed in 533ish. How
many threads when it crashes?
The message you are replying to was written before 533! Talking about
the 529-531 builds, they got up to 550 to 600 threads before stalling.
But see below.
node refs don't necessarily help, my transient node does fine with the
same ones. An overloaded permanent node seems unable to maintain
contact, presumably some function of the overloaded status is
interfering with maintaining node references. Just another data point.
There was a problem with node references getting deleted, fixed in 535.
Build 525 seemed to regain its node references after going down to nil
for a time, in that sense is much better. 534 has run for 12 hours with
just one node reference which it has contacted 100%, so 535 *would* be a
real improvement. 535 seems much better from the reference point of
view, but I can only run it for a few hours before it runs away with
5-600 threads again and stalls (OS dependent, but I have no reason to
believe it would not reach a 10 000 thread limit!). It also shows high
CPU usage for very little useful contacts again.
--
Roger Hayter
--
Roger Hayter
_______________________________________________
support mailing list
[EMAIL PROTECTED]
http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/support
