On Mon, Jul 19, 2004 at 05:02:42PM +0200, Zenon Panoussis wrote: > > Toad wrote: > > >Strange. What is your logLevel ? > > Well, that's relative. The log level is set to debug, but the > log file is a FIFO, where a simple perl script greps for URIs > and dumps the rest. My idea was to feed those URIs to mnogosearch > and create a non-anonymous search engine fo freenet.
Okay, this is your basic problem. I don't even use logLevel=debug any more. It produces crazy amounts of data, uses a lot of CPU, and the system cannot keep up. I doubt that feeding it through a FIFO will make that much difference - much of it is probably the generation side (as well as the syscalls). I recommend you set the following: logLevelDetail=freenet.client:debug That should catch all the URIs, probably. You don't need debug log level on everything else, and having it will really mess stuff up. > > >Won't make much difference. The setting you want is maximumThreads. > > I took them down from 120 to 60, saw hardly any difference at all. > > >>I also set diagnosticsPath=/dev/null to stop the constant > >>writing on disk (is there a better way to say "disable stats, > >>I don't need them"?) and that didn't help much either. > > >You do. The node uses them for estimating load. My long-lived unstable > >node's stats dir is 3MB, as is my long-lived stable node's stats dir. > >It's not a problem. > > It's not the size of the stats on disk I want to avoid, but the > extra I/O that comes from keeping them. Which is minimal, if the OS has enough memory to batch the writes into one medium sized write every 30 seconds. > > >Set maximumThreads=60 (remove the leading %), logLevel=error, and > >doCPULoad=true (is your node pegged on CPU, or just on I/O? what's the > >idle % typically?). > > OK, I'll give it one more try to see if it's the logging that > does it. Of course, the logging is required for what I want > to do, so I don't know how to get around the problem, if that's > where it is. > > The normal load of the machine is around 0.10 an I/O is its > big problem in general. I have seen similar problems when running logLevel=debug on my own nodes. The solution is to set logLevel=error and logLevelDetails for the individual subsystems you want to monitor. -- Matthew J Toseland - [EMAIL PROTECTED] Freenet Project Official Codemonkey - http://freenetproject.org/ ICTHUS - Nothing is impossible. Our Boss says so.
signature.asc
Description: Digital signature
_______________________________________________ Support mailing list [EMAIL PROTECTED] http://news.gmane.org/gmane.network.freenet.support Unsubscribe at http://dodo.freenetproject.org/cgi-bin/mailman/listinfo/support Or mailto:[EMAIL PROTECTED]