Backoff reasons over a long period from my node: 5289 ForwardRejectedOverload (RO before Accepted in request) 440 ForwardRejectedOverload2 (RO after Accepted in request) 161 ForwardRejectedOverload5 (RO before Accepted in CHK insert) 11 ForwardRejectedOverload3 (RO before Accepted in SSK request) 8 ForwardRejectedOverload4 (RO after Accepted in SSK request) 7 ForwardRejectedOverload6 (RO after Accepted in CHK insert)
1475 FatalTimeout (Timeout in request) 914 AcceptedTimeout (Timeout waiting for Accepted in request) 77 AfterInsertAcceptedTimeout2 (Timeout after accepted in CHK insert) 28 AfterInsertAcceptedTimeout (Timeout after accepted in SSK insert) 13 Timeout3 (Timeout waiting for Accepted in CHK insert) Thus, 63% of all backoffs are caused by requests being RejectedOverload'ed (presumably pre-emptively) before receiving an Accepted, in a request. Pre-emptive request rejection was instated to prevent the load limiting system having to rely on timeouts (as obviously timeouts are bad). We reject requests when the average ping time is high (700-1500ms), when the bandwidth limiter is delaying blocks for too long (1000-2000ms), and when we calculate that we don't have the bandwidth. The second should be obsolete, because of the third. It isn't. Why isn't it? I will investigate... -- Matthew J Toseland - toad at amphibian.dyndns.org Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so. -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 189 bytes Desc: Digital signature URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20060819/7da7e189/attachment.pgp>
