From Bugzilla Helper: User-Agent: Mozilla/4.76 [en] (X11; U; Linux 2.2.18-cdhs i686) Description of problem: We use two dhcpd servers in our network, for high availabiliy. Or at least we thought so. When client starts its networking, pump chooses one of the servers. Now if we stop dhcpd on one of the servers, all clients that chose that one will loose their network after what pump thinks it's max lease time. (another bug, pump doesnt care about default and max lease time defined in server's dhcpd.conf) How reproducible: Always Steps to Reproduce: 1. have a network with at least two dhcp servers 2. have some clients; check that some clients got their IP info from one and others from the other box. 3. stop one of the dhcp servers. Actual Results: try pump -R on clients. those that got respone from still running server return as expected, others, that talked to now stopped server, say Operation failed. Expected Results: Even if rfc says that renewal request is unicast, I think it would be wise in such situation to send a broadcast again to see if there are some other servers available and not to fail quietly. Additional info: We use strict mac address to IP mapping, no free IPs floating around. Same dhcpd.conf on both servers, which are RedHat 6.2 (dhcp-2.0-5). Clients are RedHat 7.0 (pump-0.8.3-2).
closing as supported products don't use pump anymore