Hi, In the meantime, things are getting much worse here:
http://freenet.thisisatest.de/part_of_wrapper.log-2 As you can see, the node has started restarting on 2006/11/02 11:51:09, and still (2006/11/04 13:44:17) hasn't finished yet. That's 48 hours and counting. Guido Am Mittwoch, 1. November 2006 18:45 schrieb toad: > Okay, the long start-up times are due to a hack we incorporated to not > clobber everyone's cache during the migration to a split store/cache. I > will remove this very soon and make the store and the cache equal sizes. > This will also prevent the store+cache temporarily using more than the > specified limit. > > On Wed, Nov 01, 2006 at 05:07:35PM +0100, Guido Winkelmann wrote: > > Am Samstag, 28. Oktober 2006 17:43 schrieb Guido Winkelmann: > > > Hi, > > > > > > Am Donnerstag, 26. Oktober 2006 22:53 schrieben Sie: > > > [...] > > > > > > > Bizarre. > > > > > > > > The first time, it takes from 00:13:10 to 01:14:15 *the next day* (so > > > > around 25 hours) to read half of the store. The second time it takes > > > > from 01:18:04 to 03:11:33 (2 hours) to read the entire store. There's > > > > something else going on here, no idea what. Does this happen often? > > > > How's memory usage? > > > > > > The problem persists. Right now (17:30) the node is at "Reading store > > > prior to shrink: 34%", the startup of the node was at 02:09 in the > > > morning. > > > > > > There is nothing special running on that box (besides freenet), a few > > > standard services (a webserver, a mailserver, a small postgres > > > database...), but that's it. Memory usage is also pretty standard, with > > > about 66 MiB of swap in use (which is nothing special fo a server > > > machine.) Everything else on the machine is running normally, I don't > > > even see any significant slowdown anywhere. (Even though freenet is > > > using 99% CPU.) > > > > > > Anyone else got this problem? > > > > > > Guido > > > > Another follow-up: > > > > Right now, I'm mostly unable to use freenet, because most of the times I > > find my node "in the process of starting up". > > > > This might be related: > > During one of the rare occassions when my node was actually up and > > running, I found it to be noticeably slower than usual (even for > > freenet). Bandwidth usage was about 2 KB/sec (in+out), of 100 KB/sec > > allowed (about a tenth of the usual amount). Also, the java process was > > still using 99% CPU constantly. > > > > Guido -- Mailing list archives are not a proper form of documentation.