> My node has gotten well connected in the Freenet work, pushing out data > at max bandwidth with a few thousand local queries per hour. But after > 1-5 hours performance collapses: routing time becomes a few thousand > milliseconds, all incoming requests are rejected, but no bytes are sent > and the number of inbound connections is much higher than the number of > outbound connections. There's hardly any network, cpu or disk activity. > The log shows some debug exceptions at CPAlgoRoutingTable.isRoutable and > a lot of out of memory errors "while receiving message" - even though > according to the Environment page the JVM has 10's of unused Mbs.
But what happens when the node tries to allocate 11 megs for use in a temporary buffer... Most applications, including fred, isn't very good at recovering from hitting a memory roof once. Increase the maximum memory allowed for the node and the problem will probably go away. 128 megs with a 10Gig storage and a well integrated node is to close to what the node might need during some circumstances. > > > Questions/remarks: > > 1) Does anyone have a node that actively runs for a number of days > unattended without major problems? If so, I'd be interested to know your > setup. Yes, I have had nodes up for both days and weeks. My setup is basically nothing special, most settings are left to default values. I do, however, allow my nodes to use significantly more memory though (not that they really do use it but still). /N _______________________________________________ devl mailing list devl at freenetproject.org http://hawk.freenetproject.org:8080/cgi-bin/mailman/listinfo/devl
