-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Oskar Sandberg wrote: | On Tue, Oct 29, 2002 at 11:53:45AM -0500, Gianni Johansson wrote: | |>The new load balancing code is a very good thing indeed, but I doubt it will |>save individual nodes from massive overload.
It didn't save me.. |> |>Has anyone conclusively figured out the CPU usage issue? | | | I had high CPU usage yesterday when the node was overloaded, but nothing | to indicate runaway 100% usage. The dump you sent me didn't contain any | giveaways either. I think I had 3 threads which used more cpu than the average. They used together about 30% of my CPU (~300 MHz) and that was 300% more than other freenet threads. CPU usage was constant so I suppose it was what you were looking for. I was using qthreads at the time. Unfortunately I didn't check from diagnostics where these threads were used (maybe I couldn't have identified them from there anyway so it's probably not a big loss). After that I switched back to 500 series. |>I suspect that there might have been a regression somewhere between 600 and |>603, however, it is impossible to test on the public network under the |>present conditions. Its even impossible to use it for anything right now. I guess the network got many many new _transient_ nodes after the release of 0.5 and this put the net on to its knees. This means every nontransient node got a lot more work to do. It didn't help at all that nontransient node keepers stopped their node at the time since CPU usage just went skyhigh and they couldn't use their computer any more. It just caused more load to other nodes. Node refs dropped from 80 to 40 yesterday and it has slowly climbed to current 50 (of which 30 are constantly backed off) Many of them, who tried freenet, are left now again but my node keeps still pushing at its limits: * localRequestsPerHour: 27435.0 |>Finally what about rate limiting Announcement requests? Aren't announcements |>quite expensive for the node to process? | | | It's probably not that significant. | I remember having about 13 threads at maximum dedicated to announcements according to environment information yesterday. Currently my overloaded node has the following thread usage (configured total: 200): Total pooled threads 229 Available pooled threads 33 Pooled threads in use 196 freenet.node.states.announcement.NoReply 4 freenet.node.states.data.DataStateInitiator 1 freenet.node.states.announcement.NoExecute 2 freenet.interfaces.LocalInterface$ConnectionShell 1 freenet.Message: NodeAnnouncement 2 freenet.interfaces.PublicInterface$ConnectionShell 116 freenet.OpenConnectionManager$ConnectionJob 51 freenet.Message: QueryRestarted 5 freenet.Message: DataRequest 6 freenet.Message: QueryRejected 8 -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.0.7 (GNU/Linux) Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org iD8DBQE9vvrBYYWM2XTSwX0RAnhuAJ9ne6TEF8bM7ZgPgf1/fiPI61pLAwCfcRr+ sZLXLuZdpAXcCAJMi9f11nU= =8Ezq -----END PGP SIGNATURE----- _______________________________________________ devl mailing list [EMAIL PROTECTED] http://hawk.freenetproject.org/cgi-bin/mailman/listinfo/devl
