I have the following situation: Centos7, 2 NIC with following setup:
em2(NIC2) :
82.79.24.74/29 82.79.24.75/29
em3(NIC3) :
192.168.0.240/24 192.168.133.240/24
NIC1 and NIC4 are not used. The em2 ips are public ips and em3 ips are private ips. The asymmetrically routed (outgoing routes and incoming routes are different) packets are accepted, the rp_filter = 2 is set. The firewall is disabled too, there is no active firewall.
Everything works fine, except one. Sometimes the host(kernel) randomly ignores a random IP for no reason, for example one from my clients IP and they cannot access nothing from the server. They cannot either ping the server.
14:18:16.553360 IP 81.12.176.198 > 82.79.24.74: ICMP echo request, id 1, seq 2171, length 40
14:18:21.548878 IP 81.12.176.198 > 82.79.24.74: ICMP echo request, id 1, seq 2172, length 40
14:18:26.547637 IP 81.12.176.198 > 82.79.24.74: ICMP echo request, id 1, seq 2173, length 40
14:18:31.547129 IP 81.12.176.198 > 82.79.24.74: ICMP echo request, id 1, seq 2174, length 40
I've captured the ICMP packets, and the packets are arriving, but there isn't any reply packet on any interface(em2, em2, lo). I can ping the client's IP from the host:
Could anybody give some advice where can I start the researching and locating the source of the problem?