I suppose those are all spoofed udp packets as they were the last time I checked them :(

I do not have direct access to the upstream links so I cannot trace them and link them to a specific bw supplier :(

just increased the limits but still getting the drop rule hit hard... it's difficult to justify these spikes as legit traffic..

check from 23:21 onward
http://pastebin.com/jUjzyKY6

I do not think I'm the only one in this situation as I saw many people discussing these problems recently :/

Il 07/01/2011 23:27, Christoffer Pedersen ha scritto:
No offense, but have you tried to look at where those dos attack comes from? 
You could block the IP-address of the attacker.

/Chris

Den 07/01/2011 kl. 22.32 skrev Marco Padovan:

I thoutgh about that too... but monitoring the situation closely it appear to 
be cristal clear:

http://pastebin.com/asHm8GkW

I getting spikes of 50k packets in very short periods (<60seconds)

I'll try to monitor all my servers in HLSW seeing how much time they are going 
offline...
btw... seeing the spikes were that big I think I can increase the limit a 
lot... maybe 25 :)

Il 07/01/2011 22:22, frostschutz ha scritto:
On Fri, Jan 07, 2011 at 08:09:40PM +0100, Marco Padovan wrote:
20 minutes later:
Chain QUERYLIMIT (4 references)
      pkts      bytes target     prot opt in     out   source               
destination
    396253 20611768 ACCEPT     all  --  *      *       0.0.0.0/0            
0.0.0.0/0           limit: avg 15/sec burst 5 mode dstport
     50483  2675483 DROP       all  --  *      *       0.0.0.0/0            
0.0.0.0/0
If the number of dropped packets keeps rising slowly here,
you are probably dropping legitimate queries. Maybe the limit
is a bit too low then. Also consider using a larger burst.
The burst will allow short, random spikes, but under actual
and constant DoS, the limit will still be respected, same as
without burst.

I'd try limit 20 burst 40 here and see how that goes. You can
be generous with burst as it will vanish completely during
a DoS attack anyhow (and it will take 40 below-limit seconds
to recharge).

another box of ours that generally suffer a lot of is now reporting:

Chain QUERYLIMIT (4 references)
      pkts      bytes target     prot opt in     out     source               
destination
    333352 16966756 ACCEPT     all  --  *      *       0.0.0.0/0            
0.0.0.0/0           limit: avg 15/sec burst 5 mode dstport
    563098 29844034 DROP       all  --  *      *       0.0.0.0/0            
0.0.0.0/0
drop>>   accept is to be expected during a DoS attack.

nobody complained yet... so looks like its holding :)
Test it yourself - see if you can get a complete server
list using the standard steam server browser. If half
of your servers are missing there most of the time
(while there is NO DoS going on), chances are your
limit is too low.

Regards
frostschutz

_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux
_______________________________________________
To unsubscribe, edit your list preferences, or view the list archives, please 
visit:
http://list.valvesoftware.com/mailman/listinfo/hlds_linux

Reply via email to