[CentOS] database node / possible SYN flooding on port 3306

Nataraj incoming-centos at rjl.com
Sat Jul 21 15:09:22 UTC 2018

On 07/21/2018 03:48 AM, Leon Fauster via CentOS wrote:
>> Am 20.07.2018 um 18:52 schrieb Nataraj <incoming-centos at rjl.com>:
>> On 07/20/2018 03:56 AM, Leon Fauster via CentOS wrote:
>>> Hi folks,
>>> I have here a database node running
>>> # rpm -qa | grep mysql-server
>>> mysql55-mysql-server-5.5.52-1.el6.x86_64
>>> on
>>> # virt-what 
>>> vmware
>>> that seems to have a connection problem:
>>> # dmesg |grep SYN |tail -5
>>> possible SYN flooding on port 3306. Sending cookies.
>>> possible SYN flooding on port 3306. Sending cookies.
>>> possible SYN flooding on port 3306. Sending cookies.
>>> possible SYN flooding on port 3306. Sending cookies.
>>> possible SYN flooding on port 3306. Sending cookies.
>>> I adapted already following:
>>> # sysctl -a |grep -E 'maxconn|syn_backlog'
>>> net.core.somaxconn = 2048
>>> net.ipv4.tcp_max_syn_backlog = 2048
>>> but ListenOverflows and ListenDrops values are still high
>>> # cat /proc/net/netstat | awk '{print $21 "-" $22 }'
>>> ListenOverflows-ListenDrops
>>> 13568-13568
>>> any suggestion? 
>> Use tools like tcpdump/wireshark  and further examination of logfiles to
>> determine where your attack is coming from, i.e. single IP address or
>> multiple ip addresses (BOT attack).
>> If attack is impairing your Internet service, contact your ISP.  Most
>> decent ISP's should deal with this situation for you.disabled
>> If attack is not impairing your service and you choose to deal with it
>> yourself, then, if from a fixed IP address block that IP from your
>> firewall if you have one, otherwise, use IPtables on the server.  If
>> your having bot attacks, or blocking attack causes source IP address to
>> be changed, then look at fail2ban.  Basically you want to configure
>> fail2ban to limit the number of requests per unit of time and block IPs
>> that exceed that.  Also, consider weather your database needs to be
>> publicly accessible from the Internet.
> Actually the database node is a backend system in a private network, so "all" 
> traffic is legitimate. The main traffic comes from the web node (cms/php). Resources 
> of the db node seems all to be okay (cpu/mem/load). So i do not see any bottleneck ... 
> --
> LF
> https://www.google.com/search?client=ubuntu&channel=fs&q=mysql+possible+syn+flooding&ie=utf-8&oe=utf-8
> _______________________________________________
> CentOS mailing list
> CentOS at centos.org
> https://lists.centos.org/mailman/listinfo/centos

I would still use tcpdump and/or wireshark to find out more what is
going on and if all the incoming connections seem correct, you may have
to tune some parameters to allow a greater number of connections.  Also,
if you have iptables connection tracking turned on, this might need to
be bypassed for connection on your database port.  Connection tracking
does not always work well for large numbers of connections.  But check
to see weather one or more clients are somehow failing to connect and
retrying the connection at a high rate for some reason.

Also Try

and maybe this:



More information about the CentOS mailing list