I'm hoping that this might shed a bit more light on why we are having issues bootstrapping nodes. I've had to temporarily disable my seed node. I am the senor network engineer for a small co-location facility.
Over the weekend I started to get reports that the bandwidth throughput for was dropping off. After I shutdown the seed node things went back to normal. We have several customers that where effected, with really long load times. I have over 20 Mbits allocated to Freenet, and it's running on its own server. Prior to shutting down the node I cut and pasted some of the stuff from the stats page. Also I have the wrapper log. I also have a cut and paste of the nat table on main router. I don't think that this was an attack nor do I think that it was a bandwidth issue. However I DO think that it was overflowing the Nat table. I am taking steps to resolve this by giving the Freenet seed node a public ip and skipping nat. I will resend the node ref later today after I take care of this. However while this will take care of my problem and make the seed node not overload nat, I don't think that we have fixed the real issue with bootstrapping new nodes. It didn't become an issue until sometime Thursday/Friday. I would like to dig into this issue some more, but I am unable to do so for the next few days. > -----Original Message----- > From: devl-bounces at freenetproject.org [mailto:devl- > bounces at freenetproject.org] On Behalf Of Matthew Toseland > Sent: Saturday, February 05, 2011 1:40 PM > To: devl at freenetproject.org; support at freenetproject.org > Subject: [freenet-dev] Call for seednodes and explanation of current > problems > > We need more seednodes. I will explain the broader situation below. If > you can run a seednode - which means you need a forwarded port, a > reasonably static IP address (or dyndns name), and a reasonable amount > of bandwidth (especially upstream), and a reasonably stable node, > please send me your opennet noderef (from the strangers page in > advanced mode), and enable "Be a seednode" in the advanced config. > Thanks. > > Details: > > One of the problems Freenet has at the moment is that bootstrapping a > new node can take an awfully long time - 20 minutes or more sometimes. > It is not clear why; we seem to either get rejected by seednodes (most > of the time), or they return nothing, maybe a few "not wanted" notices, > or they return lots of noderefs and we manage to announce. > > This might be due to bugs. 1343 fixed a bug that apparently badly > affected some seednodes. However it appears most seednodes have > upgraded now. > > There doesn't seem to be a problem with losing connections - backoff > yes but once a node is connected it seems to mostly stay connected. > > The most likely answer seems to be that we just don't have enough > seednodes to cope with the load. > > It is also possible that this is due to an attack. It did come on > relatively suddenly a few weeks ago (it was bad before but it got much > worse), and it seems to have got significantly worse in the last week. > It is not clear how we would identify an attack if that was the > problem; there are no obvious signs so far. > > It is also possible it is a client-side bug. Testing of the master > branch would be useful, it has some small changes.
