[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: (FWD) Results of stress testing the current version of the Onion Router



I killed all the httpap's and restarted a new one and it still did not work :(

I had to kill off both the httpap and the OR on the first machine and then 
restart them both before it would start working again. However, I did not 
have to restart the other two onion router processes on the other two 
machines. No messages were logged for either the httpap or the OR 
processes. Logging was set at ERR for everthing.

However, I suspect that the output buffers may not have been flushed on the 
processes that hung. Does anyone know how to make a process not buffer its 
output so that it gets written immediately?

I monitored the progress of the test carefully, and things didn't start 
going to hell until there were 64 simultaneous connections. The throughput 
was about 35Mps up until that point. After 64 simultaneous connections 
things locked up and all subsequent connections would fail.

At 12:29 AM 8/30/2002 -0400, you wrote:
>On Thu, Aug 29, 2002 at 08:13:44AM -0400, Bruce Montrose wrote:
> > By down I mean that the httpap was not talking to anybody, so i suppose 
> the
> > problem may be confined to that process and the actual "or" processes may
> > have weathered the storm.
>
>There's a good chance that's true -- httpap has a MaxConn option (500
>in the config file in cvs) that says "reject the connection if there
>are too many httpap processes currently".
>
>Next time you get into this situation, killall the httpap's, start a
>new one, and see if things are working again. (If so, then wait til we
>get around to making Privoxy work. :)
>
>--Roger