On Thu, Feb 06, 2003 at 12:14:13PM -0700, Jack D.Ripper wrote:
> Class Threads used
> Checkpoint: Fetching ARK freenet:SSK 34
> freenet.interfaces.LocalInterface$ConnectionShell 5
> freenet.OpenConnectionManager$ConnectionJob
>
> This is about thirty minutes after I overwhelmed my node by trying out Frost,
> and maxed out my process table. Build Number 554 CVS Revision 1.90.2.30
Hmmm.
>
> It appears that Fred is trying to catch up on all those ARK jobs that didn't
> get fired off because of being unable to spawn threads, and as a result
> recovery from an overload is taking a *LONG* time.
>
> Is there some configuration option I can use to a) help prevent overloads, and
> b) help recover from them faster? My main limitation is CPU, not memory. I
> have plenty of memory, but only a Pentium 166MMX CPU. Upgrading the CPU on my
> FreeNetmachine is not practical (I'd basically need a new machine). I
> already bumped my max connections WAY down (to 10), which keeps my node from
> getting overloaded in normal use (i.e., just fetching stuff via port ),
> but Frost seems to utterly overwhelm it.
maxConnections is ignored. maximumThreads is the relevant parameter.
Try:
maximumThreads=60
defaultResetProbability=0.01
>
> - JDR
--
msg02829/pgp0.pgp
Description: PGP signature