On Friday 30 November 2007 03:43, Michael Rogers wrote: > Matthew Toseland wrote: > > In 1076, I made a perhaps fortuitous mistake: I changed RequestStarter to feed > > local requests through the same selection process that remote ones go through > > (which is good), but I accidentally didn't put in a delay. So once there was > > a local request queued, RequestStarter would busy-loop until it was able to > > send it. > > I started a large insert and this happened: > > Bandwidth > > * Input Rate: 47.8 KiB/sec (of 48.0 KiB) > * Output Rate: 107 KiB/sec (of 12.0 KiB) > * Total Input: 230 MiB (11.2 KiB/sec) > * Total Output: 426 MiB (20.7 KiB/sec) > * Payload Output: 32.1 MiB (1.56 KiB/sec)(7%) > > I had to shut down my node in order to send this email because TCP > connections were timing out...
As I said, that's a separate bug, which hasn't been fixed yet as it's hard to reproduce, and which definitely happened in 1075 so is not a result of the busyloop bug. The critical thing here is not only is the output rate way over the limit, the payload output % is very low. This is a low level problem of some kind: excessive retransmission perhaps. Note also that it doesn't happen for me. :< Anyway if you could get logs with logleveldetails=freenet.node:MINOR, that would help. > > Cheers, > Michael -------------- next part -------------- A non-text attachment was scrubbed... Name: not available Type: application/pgp-signature Size: 189 bytes Desc: not available URL: <https://emu.freenetproject.org/pipermail/devl/attachments/20071130/32a2dffd/attachment.pgp>
