ipconntrack suddenly became toooo large
In your syslog do you see:
ip_conntrack: table full, dropping packet.
If so, your conntrack table is full. You can also check the current number of connections tracked using this command:
username@localmachine:~$ cat /proc/slabinfo|grep conntrack
nf_conntrack_expect 0 0 248 16 1 : tunables 0 0 0 : slabdata 0 0 0
nf_conntrack_ffffffff81d11da0 1499 1536 320 12 1 : tunables 0 0 0 : slabdata 128 128 0
The first/second number is the value that the system is currently tracking (active/total). You can check the current max number of connections your kernel can watch using this command:
sysctl net.netfilter.nf_conntrack_max
For CentOS/Redhat, it is called ip_conntrack. If you don't see nf_conntrack_max check this:
sysctl net.ipv4.netfilter.ip_conntrack_max
You can increase using the sysctl -w command:
sysctl -w net.netfilter.nf_conntrack_max=131072
For example. The number is dependent on how much memory you have in the machine and whether the kernel can store the connection informations. The above setting doubles the default Ubuntu setting of 65536.
When you find a good value, add it to your /etc/sysctl.conf file to apply it on reboots:
# increase the number of conntrack max
net.netfilter.nf_conntrack_max=131072
As long as your machine and the upstream network equipment can handle the traffic, you should be good. As you had mentioned it is good to see what caused this increase in traffic. If you are being DDOSed you should try to figure out an anti-DDOS solution which could involve a combination of a software firewall or an appliance depending on your application and financial motivations.