Every four hours after server reboot TCP connections fail to establish in both directions. The problem lasts several minutes then gone. No events in syslog seem to be relative, even in dmesg. ICMP ping is still OK at the same time, even with -s 1500 (mtu is 1500). I presumed that some process opens too much connections, and set up monitoring script with ss -at4 every ten seconds, but found nothing, no new suspicious processes and connections appear. Distro is Debian Stretch-like. Links are bonded. What can I check to solve it? UPD. I found that 4 hour period counts from restart of networking.service, not just reboot itself. Also it is worth mentioning that network interfaces are connected over OVS on this host. TCP kernel options and limits are intact and identical with other nodes.
Asked
Active
Viewed 156 times
0
-
Check your logs/journal for relevant information (perhaps consider enabling auditing) to see what happens every 4 hours. – user9517 Sep 06 '22 at 08:56
1 Answers
0
I don't know why it was happening with this exact time interval but I finally discovered that I have same IP in the network. It's a surprising thing for me that system doesn't signal about identical IP in its network, as Windows does
UPD The culprit was a virtual interface on Cisco device in the network. Cisco has 14400 seconds default ARP timeout (four hours).
Макс Кринж
- 13
- 3