Thus spake Olaf Selke (olaf.selke@xxxxxxxxxxxx): > Olaf Selke schrieb: > > > > in the last weeks bandwidthrate gave only few control about the number > > of connections. To check this I've just reduced BandwidthRate from 16000 > > to 1000 KB. Let's wait one or two days and see if connections and > > connection rate considerably drops. We'll see... > > good morning world, > > after two day running with a BandwidthRate of 1000 KB there're still 17k > tcp connections on my router. Reducing bandwidth by more than a factor > of ten only reduces the number of connections less than a factor of two. > On the other hand the number of new connections per second noticeably > dropped by about 90%. Interesting. Lot of long-lived streams+circuits must be your primary cause of (memory?) load. Has the CPU usage and other issues you were noticing reduced? Was it just the out of memory error that was your primary problem? Has the memory usage improved? It looks like you are not currently in the consensus. Did you just restart the node? It will be interesting to see what bandwidth our bandwidth authorities give you, now that they are running again. Is the BandwidthRate still 1k? Can you leave it there for just another day? I understand your frustration with the change in behavior of the config option. But with the right experiments, we can figure out the right way to fix this bug. I am still not convinced hacking the MaxAdvertizedBandwidth is the right nob to tune to fix the issue. Maybe it is the best we can do in the short term, but I'd still like to know more first. Thanks a lot for your help, -- Mike Perry Mad Computer Scientist fscked.org evil labs
Attachment:
pgp0BpYv6PgcN.pgp
Description: PGP signature