Loss of Network Adress is DHCP Server failed for some hours

Dear NM-experts,

we are using CentOS 7 with NetworkManager out of the box, using DHCP reserved adresses also for our servers. 
Last night, our DHCP server (+ backup servers, and quite a bit more infrastructure) failed for about 15 
Since our lease time was set to 4 hours, this meant several leases expired in that timescale. 

While everything is recovered now, all CentOS 7 machines are STILL without IP address. Looking at the log 
files, I see from NM: 
scheduling DHCPv4 restart in 120 seconds, 3 tries left (reason: lease expired)
request timed out
scheduling DHCPv4 restart in 120 seconds, 2 tries left (reason: lease expired)
until finally the three tries have passed. Since dhclient is now not running anymore, and NM does not care 
about the interface anymore, the machine stays in a broken state. 

Restarting the machines, or network manager, resolve the broken state. 

This is very much against my expectations from older systems not yet using NM. Native dhclient just keeps 
trying with some large interval, so the system recovers once the server is back. 
I think the same is true for native dhcpcd, and also for any more simple device I know (Android, Switches, 
BMCs etc.). 

Of course, a generally good idea would be to increase the lease time. 
However, I still would like to force NM to behave like a "normal" non-intelligent DHCP client, i.e. just 
"keep trying". 

Is this a bug, or a feature? 
How can I configure the behaviour to be the expected one (apart from reverting to a non-NM setup)? 


[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]