I'm impressed!

Using command line mail to send email ~ 287 kb (HTML format with embedded
images)
Time interval about 7 minutes
Email received (and process with custom mailet): about 1000
Total mailbox size during this 7 minutes: ~ 287 MB

Machine:
James Server running on AMD64 3500MHz running FC4 x86_64 with 1GB RAM. JVM
-Xms512m -Xmx768m

spam machine: running AMD K6-2 400 MHz

Machine and spammer are on the same 100 BaseT

Maybe I should try running the spam machine on P4 and see how it goes. But
for now, I am satisfied that James holds very well and it doesn't crash at
all.



On 1/21/06, Stefano Bagnara <[EMAIL PROTECTED]> wrote:
>
> Edward Tan wrote:
> > Hi,
> >
> > I am given task to stress test/hack our development before we deploy.
> Just
> > now I experience a problem with JVM runs out file descriptors because it
> > tries to open so many files/socket. This is not the problem of the OS or
> JVM
> > or Tomcat. It just that some portion of the software does not have
> > self-limiting/queueing facility.
> >
> > It gives me an idea, if I know a valid email address in James server, I
> can
> > write daemon to send floods of very small emails to that mailbox. I
> notice
> > that James writes to its spools, mailboxes, etc. Will it be possible to
> kill
> > James using this technique because James suddenly tries to open so many
> > files at almost the same time?
> >
> > Or is there any queueing facility. It will be fair enough that if there
> is
> > flood of emails, the other SMTP party will get transmission failure. But
> not
> > in the situation where the recipient SMTP keeps receiving but bombs out
> > because it tries to open too many files. I believe, James does thing
> like in
> > the first case and not the later. But I may be wrong.
> >
> > Anyone can verify/give assurance that James will not crash when receive
> tons
> > of valid email to valid mailbox almost at the same time?
>
> It seems you already know how to DOS James: simply try it and let us
> know wether it crashes or not :-)
>
> Who would you trust if not yourself?
>
> You can configure the number of connections for each service, the number
> of spooling threads and more.. just try it: it should not be difficult
> to reproduce your test.
>
> Stefano
>
> > Thanks
> > Edward
> >
>
>
> ---------------------------------------------------------------------
> To unsubscribe, e-mail: [EMAIL PROTECTED]
> For additional commands, e-mail: [EMAIL PROTECTED]
>
>

Reply via email to