The following procedure was tested on Bright 9.1 on CentOS 7.9.
WARNING: Please note that an update of certain Bright packages can remove some of the changes that were done manually here. If the packages node-installer or node-installer-nfsroot are updated/reinstalled on the head node, then step 7 of this article has to be done again.
1 – Disable ntpd on the head node
Run the following commands on the head node:
# systemctl disable ntpd
# systemctl stop ntpd
# cmsh
% device use master
% services
% use ntpd
% stop
% set autostart no
% set monitored no
% commit
2 – Disable ntpd on the software images
Run the following commands on the head node (this has to be done for every software image)
# chroot /cm/images/<software images>
# systemctl disable ntpd
# exit
3 – Install chrony on the head node and software images
Run the following commands on the head node:
# yum install chrony
# yum –installroot=/cm/images/<software image> install chrony
… repeat this last step for every software image
4 – Configure chrony on the head node
In the head node, edit the /etc/chrony.conf file and add the following line (replace the network for the network from which the compute nodes will access the head node)
allow 10.141.0.0/16
5 – Configure chrony on the software images
In every software image edit the /etc/chrony.conf file, comment all the lines starting with “server” and add a line which points to the IP address the compute nodes use to reach the head node, for example:
server 10.141.255.254 iburst
6 – Configure the head node and compute nodes to manage chronyd
Run the following commands on the head node:
# cmsh
% device use master
% services
% add chronyd
% set monitored yes
% set autostart yes
% commit
% start
% category use <category of the nodes>
% services
% add chronyd
% set monitored yes
% set autostart yes
% commit
7 – Configure node-installer to use chrony instead of ntpd
Run the following commands on the head node:
# yumdownloader chrony libseccomp
# rpm –root=/cm/node-installer -ivh chrony-*.rpm libseccomp-*.x86_64.rpm
8 – Reboot nodes to apply the changes
All the compute nodes have to be rebooted so the changes in the software images will be applied.