[PCM] CN IP loss after 12 hours due to the dhcp.lease timeout value 43200

[xyzcaoli]

1. after PCM installation under rhel 7.2, one setting is here
   [root@pcmrhe72 ldap]# tabdump site | grep lease
   "dhcplease","43200",,
2. Then add node and provision PCM compute node, after node provision, everything goes well for compute node getting and keeping IP addr
3. but after 12 hours, the PCM compute node will lost IP info, those node will lost connection from PCM Management Node
4. the xcat/OS version is below
   [root@pcmrhe72 ldap]# lsxcatd -v
   Version 2.11 (git commit c3bb309, built Tue Dec 22 23:57:00 EST 2015)
   [root@pcmrhe72 ldap]# rpm -qa | grep -i xcat
   xCAT-genesis-base-x86_64-2.9-snap201503060727.noarch
   xCAT-buildkit-2.11-snap201512222357.noarch
   perl-xCAT-2.11-snap201512222357.noarch
   conserver-xcat-8.1.16-10.x86_64
   elilo-xcat-3.14-4.noarch
   xCAT-genesis-scripts-x86_64-2.11-snap201512222357.noarch
   ipmitool-xcat-1.8.15-1.x86_64
   xCAT-client-2.11-snap201512230028.noarch
   xCAT-2.11-snap201512230028.x86_64
   xCAT-vlan-2.11-snap201512222357.noarch
   xCAT-server-2.11-snap201512230028.noarch
5. OS info
   [root@pcmrhe72 ldap]# lsb_release -a
   LSB Version: :core-4.1-amd64:core-4.1-noarch
   Distributor ID: RedHatEnterpriseServer
   Description: Red Hat Enterprise Linux Server release 7.2 (Maipo)
   Release: 7.2
   Codename: Maipo
6. The issue only could be encountered under rhel7.2, for example w/ the save PCM and xCAT, the issue can not be encountered SLES11.3, also the setting on site table is "43200". Seems under rhel7.2, when DHCPD IP leasing time, expired, the lease could not be renewed.

[xyzcaoli]

@zet809

[zet809]

It is found that the dhclient process is stopped later after the CN is rebooted after installation. The dhclient is triggered by NetworkManager service.
We have stoped the NetworkManager about reboot, and then the dhclient is terminated by the service. That is why the IP address is missing when using DHCP.
We are discussing how to fix it without affect other functions. Will update once there is a result.

[whowutwut]

@zet809 Thanks for this analysis! it makes sense.. My initial guess would have also been NetworkManager, i see many new OS using this and we disable it.

[daniceexi]

The issue has been fixed.

#966 for 2.11 branch
commit 52061dd for 2.12 branch

[daniceexi]

It has been fixed, so close it now.