After one of these updates (probably 1135), my node has restarted. That was yesterday at about 7 am, but I can't say for sure, because the point when this happened is no longer in the logs. It's close to 24 hours later now, and it still hasn't finished restarting.
The wrapper.log is full with lines like these: INFO | jvm 5 | 2008/04/11 03:25:22 | Freed block 60339 (duplicate) INFO | jvm 5 | 2008/04/11 03:25:25 | Freed block 60395 (duplicate) INFO | jvm 5 | 2008/04/11 03:25:27 | Key 60416/460332 OK (460 dupes, 35 failures) Am Donnerstag, 10. April 2008 17:50:53 schrieb Matthew Toseland: > Freenet 0.7 build 1136 is now available. Please upgrade. I forgot to > announce 1135; a big part of 1136 is fixing bugs in 1135. Apologies for the > recent node crashes... > Changes: > 1135: > - Increase the redundancy on splitfiles from 150% to 200%. Your node only > needs to download 100% of this to complete the fetch. The practical effect > is that inserts are 33% slower, and hopefully data persistence will be > significantly improved. This is partly inspired by Wuala, which uses 500% > splitfile redundancy. :) > - Reset node locations less often. Hopefully this will reduce location > churn and therefore improve data persistence, without causing more keyspace > clustering as we saw last year. > - Allow an opennet peer 10 minutes after disconnecting for it to reconnect, > instead of 5. Many computers take that long to reboot! > - Don't forget to disconnect from seednodes once we've finished with them. > - Add Identifier's to most messages in FCP that didn't have them. > - Hopefully fix changing request priorities causing requests to stall. > - Lots of minor bugfixes, some in the web interface, including the Method > Not Allowed bug. > - Record location changes in loation.log.txt. > - One bugfix to location swapping (we were resetting our location if we > swapped with a nearby node). > - Various internal improvements - comments, synchronization, refactoring > etc. 1136: > - 1135 would insert splitfiles with wrong metadata, resulting in the node > crashing. This is fixed, and we can even download files inserted by 1135. > - Prevent old nodes which are unable or unwilling to update from spamming > the seednodes forever. > - Default to 10% of available disk space in the first time wizard datastore > size selection page, if we are running java 1.6. > - Minor bugfixes and internal changes (mostly datastore refactoring). -- Oh my science!
