Miguel Barbosa Gonçalves wrote:

The first 5 belong to the same entity. They queried my server 9255 times in
292 seconds. This is 31 queries per second! The first IP address contacted
my server 6.79 times per second.

I find this bad, very bad in fact. So, I added a rule in my firewall so
that there only can be one state per IP address. The UDP timeout in my
firewall is around 30 seconds so this limits the clients to one connection
every 30 seconds.



There are different kinds of abusers.  One possible cause for such query rates 
is that these
are in fact NAT routers that serve a large number of systems, each making their 
NTP queries
at an acceptable rate but together sending traffic at such high rates.  Of 
course the admin
should setup a local NTP server, sync that to the pool, and refer the internal 
clients to that
server.  But admins think that is too much work, the way they do it now "just 
works".
Of course it would be less of a problem when there would be a large number of 
servers in
the .pt area.

Another kind is the jerk that just want to break things.  With the size of 
internet today, there
always are a large number of jerks even when it is only a small percentage of 
users.  Not
much that can be done about it, the jerk does not even have to be at the 
location the whois
points to because there are so many incompetent ISPs out there that do not 
perform source
address filtering (BCP 38).  They may be trying to DDOS the people that you 
think are the
cause.

A third kind is the broken NTP client.  Unfortunately for you, there are 
clients that increase
the query rate when they get little or no response.  So, when you rate limit 
the requests,
the rate actually *increases* to make up for it.  So be very careful when doing 
rate limiting
and always monitor the effect of it.

Rob



_______________________________________________
pool mailing list
[email protected]
http://lists.ntp.org/listinfo/pool

Reply via email to