I was trying to do what the article at http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.performance.html#conn... http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.performance.html#conntrack_filling_tables%3C/blockquote%3E%3C/div%3E suggested My iptables rules are ------------------------------------------------------------------------ #that's what the mentioned article suggested..I'm not sure it's working! *raw -A PREROUTING -p tcp -m tcp --dport 80 -j NOTRACK COMMIT *filter -A INPUT -i lo -j ACCEPT -A INPUT -p icmp --icmp-type any -j ACCEPT #no tracking needed for this -A INPUT -p tcp --dport 80 -j ACCEPT #that would be another question but I can't get rid of this while using ssh tunneling -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT #ssh port -A INPUT -p tcp --dport 12345 -j ACCEPT #my ip -A INPUT -s 123.123.123.123 -j ACCEPT -A INPUT -j DROP -A FORWARD -j DROP COMMIT ------------------------------------------------------------------------ I keep getting these messages on my kernel log ------------------------------------------------------------------------ Apr 13 20:00:41 server kernel: ip_conntrack: table full, dropping packet. Apr 15 14:23:29 server kernel: ip_conntrack: table full, dropping packet. Apr 15 20:19:04 server last message repeated 2 times Apr 16 13:53:58 server kernel: ip_conntrack: table full, dropping packet. Apr 17 19:05:32 server last message repeated 3 times Apr 17 21:20:43 server kernel: ip_conntrack: table full, dropping packet. ------------------------------------------------------------------------ is there a way to completely disable ip_conntrack ?
John R Pierce wrote:
Masry Alex wrote:
is there a way to completely disable ip_conntrack ?
without connection tracking, NAT simply won't work.
With recent kernels, it is possible to do 1:1 NAT (mapping one private address to exactly one public IP alias on the external interface) without netfilter, but using iproute instead. It will not work for other kinds of NAT, only for 1:1 mapping.
I forgot the details, but you'll have to build and install the most recent stable kernel, and probably also update the iproute and iptables packages to the most recent stable releases. And then you can do 1:1 NAT with the ip utility. Because NAT is not activated in netfilter, ip_conntrack is not required.
On Friday 18 April 2008 12:23, Masry Alex wrote:
#that's what the mentioned article suggested..I'm not sure it's working! *raw -A PREROUTING -p tcp -m tcp --dport 80 -j NOTRACK
Do you have a chain called NOTRACK? What is setup under it?
COMMIT *filter -A INPUT -i lo -j ACCEPT -A INPUT -p icmp --icmp-type any -j ACCEPT #no tracking needed for this -A INPUT -p tcp --dport 80 -j ACCEPT #that would be another question but I can't get rid of this while using ssh tunneling -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT
OK, here is your problem. The above line should be the first line in your INPUT statement. IPTABLES reads top down so it executes the rules in the order they are placed. Since you have '--dport 80' rule before the 'ESTABLISHED,RELATED' rule it add the address to the conntrack. Every packet is being added to the conntrack making a bunch of tracking tracking the same host. If 'ESTABLISHED,RELATED' were first it would check to see if the host has already connected and allow them to continue to connect without adding then to the tracking table every time a packet comes.
You want 'ESTABLISHED,RELATED first in all your rule chains. There is a way around this if you want '--dport 80' before the 'ESTABLISHED,RELATED' and that would be like this:
-A INPUT -p tcp --dport 80 -m state --state NEW -j ACCEPT
Your rules are a mix of Stateful and Non-Stateful chose one or the other. Preferable Stateful.
On Fri, Apr 18, 2008 at 10:23 AM, Masry Alex masryalex@gmail.com wrote:
I was trying to do what the article at http://www.austintek.com/LVS/LVS-HOWTO/HOWTO/LVS-HOWTO.performance.html#conn... suggested My iptables rules are ________________________________ #that's what the mentioned article suggested..I'm not sure it's working! *raw -A PREROUTING -p tcp -m tcp --dport 80 -j NOTRACK COMMIT *filter -A INPUT -i lo -j ACCEPT -A INPUT -p icmp --icmp-type any -j ACCEPT #no tracking needed for this -A INPUT -p tcp --dport 80 -j ACCEPT #that would be another question but I can't get rid of this while using ssh tunneling -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT #ssh port -A INPUT -p tcp --dport 12345 -j ACCEPT #my ip -A INPUT -s 123.123.123.123 -j ACCEPT -A INPUT -j DROP -A FORWARD -j DROP COMMIT ________________________________ I keep getting these messages on my kernel log ________________________________ Apr 13 20:00:41 server kernel: ip_conntrack: table full, dropping packet. Apr 15 14:23:29 server kernel: ip_conntrack: table full, dropping packet. Apr 15 20:19:04 server last message repeated 2 times Apr 16 13:53:58 server kernel: ip_conntrack: table full, dropping packet. Apr 17 19:05:32 server last message repeated 3 times Apr 17 21:20:43 server kernel: ip_conntrack: table full, dropping packet. ________________________________ is there a way to completely disable ip_conntrack ?
The first question is why is your box running out of connections? A machine with 512MB can have 32768 connections in the table. If you have more than that.. you are probably getting DOS'd or something else. What state are the connections in? Does having SYN cookies help
It used to be that the only way to do that is to remove all NEW,RELATED,ESTABLISHED, etc lines in your firewall and then make sure that the conntrack and state modules didn't get loaded. I am not familiar with the preroute rule you have so I can't say if its going to help or not. Its usually easier to make the ip_conntrack table bigger or figure out why the system is getting dossed.
http://www.wallfire.org/misc/netfilter_conntrack_perf.txt
#that's what the mentioned article suggested..I'm not sure it's working! *filter -A INPUT -i lo -j ACCEPT -A INPUT -m state --state ESTABLISHED,RELATED -j ACCEPT -A INPUT -m state --state NEW -p icmp --icmp-type any -j ACCEPT -A INPUT -m state --state NEW -p tcp --dport 80 -j ACCEPT -A INPUT -m state --state NEW -p tcp --dport 12345 -j ACCEPT -A INPUT -m state --state NEW -s 123.123.123.123 -j ACCEPT -A INPUT -j DROP -A FORWARD -j DROP COMMIT
## ## No tracking #that's what the mentioned article suggested..I'm not sure it's working! *filter -A INPUT -i lo -j ACCEPT -A INPUT -p icmp --icmp-type any -j ACCEPT -A INPUT -p tcp --dport 80 -j ACCEPT -A INPUT -p tcp --dport 12345 -j ACCEPT -A INPUT -s 123.123.123.123 -j ACCEPT -A INPUT -j DROP -A FORWARD -j DROP COMMIT
## ## /etc/sysctl.conf ## sys.net.ipv4.netfilter.ip_conntrack_max=32768 # 512 MB sys.net.ipv4.netfilter.ip_conntrack_buckets=4096 # 512 MB
## ##
-
Florin Andrei -
John R Pierce -
Masry Alex -
Robert Spangler -
Stephen John Smoogen