DHCPd keeps reassigning IP addresses / breaks IPMI power control

Bug #2043135 reported by Alex Smola
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
MAAS
New
Undecided
Unassigned

Bug Description

This bug is probably related to https://bugs.launchpad.net/maas/+bug/2025468 which is supposedly fixed. That said, we're on Maas 3.4 RC2 current as of 11/9/2023 and the problem persists.

** Effect

After some time (typically 24h) after commissioning a machine, it reports the following error:

'''
Failed to query node's BMC - Connection timed out while performing power action. Check BMC configuration and connectivity and try again.
'''

** Cause

The DHCPd reassigns a new IP number to an existing MAC address, yet Maas doesn't keep track of this. This is likely due to a broken maas-helper script. An example of it is below (this is from dhcpd.leases):

'''
lease 192.168.29.237 {
  starts 2 2023/11/07 18:44:44;
  ends 2 2023/11/07 18:46:44;
  tstp 2 2023/11/07 18:44:20;
  tsfp 2 2023/11/07 18:49:20;
  atsfp 2 2023/11/07 18:49:20;
  cltt 2 2023/11/07 18:44:44;
  binding state free;
  hardware ethernet ac:1f:6b:b6:a4:a4;
  set vendor-class-identifier = "udhcp 1.23.1";
  client-hostname "disk-27.ipmi.canada.boson.ai";
  on expiry {
    set clhw =
       binary-to-ascii (16, 8, ":",
                        substring (hardware, 1, 6)) ;
    set clip =
       binary-to-ascii (10, 8, ".", leased-address) ;
    execute ("/usr/sbin/maas-dhcp-helper", "notify", "--action", "expiry",
        "--mac", clhw, "--ip-family", "ipv4", "--ip", clip, "--socket", "/var/lib/maas/dhcpd.sock");
  }
  on release {
    set clhw =
       binary-to-ascii (16, 8, ":",
                        substring (hardware, 1, 6)) ;
'''

After the initial (correct assignment to 192.168.29.237) later on the machine is assigned a new IP address as follows:

'''
}
lease 192.168.30.169 {
  starts 3 2023/11/08 19:45:03;
  ends 3 2023/11/08 19:55:03;
  tstp 5 2023/10/06 02:33:42;
  tsfp 3 2023/11/08 20:00:03;
  atsfp 3 2023/11/08 20:00:03;
  binding state active;
  next binding state expired;
  hardware ethernet ac:1f:6b:b6:a4:a4;
}
'''

Note that we're in a subnet 255.255.252.0 with 1024 IP numbers, hence 192.168.29.237 and 192.168.30.169 are in the same subnet for what Maas is concerned. I posted on https://discourse.maas.io/t/dhcpd-forgets-about-ipmi-ip-addresses-maas-3-4-ipmi-power-error/7582 but did not get any replies, hence cross-posting here.

Tags: dhcp
To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.