On 1/7/21 2:12 PM, Brigman, Larry wrote:
I understand that OKD 3.11 is old but we are still using it in production. Is there a flag or something during the installation that can be used to allow the nodes to work correctly with DHCP. We have found that the IP address is embedded in the certs and when the IP address of the nodes change do to DHCP the nodes go offline and are unrecoverable except by getting DHCP to provide the original IP addresses.
Static addresses were/is a requirement. But it's easily solved with DHCP - just add a host entry into dhcpd.conf for each of the node MAC addresses, and that IP is now reserved for that system only. Typically you would define a subnet that doesn't include the range where the OCP3 nodes are, and use host entries to give them addresses outside the space managed by dhcpd. But you can mix them if you want to.
Is there a solution? Does OKD 4.X suffer from the same fate/issue?
Because of the push-back we got around having to use static IPs that was all changed to use dhcp by default. The above method is one way (not the only) you can control it.
Shutting down OCP nodes down so long they lose their IP aren't really part of the standard procedure. Either set up the TTL for the address lease (perhaps use a special DHCP range/settings for your servers vs. clients) or do what OCP4 will have you do - kill/delete/remove the server when you don't need it - it's very easy to just create a new one when you need it (this is a feature OCP4 has that OCP3 doesn't - it will scale compute nodes just like you're used to scaling pods). So OCP4 can add and remove nodes based on demand. You won't have this problem at all.
-- Peter Larsen Staff Domain Architect Red Hat NA <https://www.redhat.com> 1600 International Drive plar...@redhat.com <mailto:plar...@redhat.com> M: 5407380733 <tel:5407380733>
_______________________________________________ users mailing list users@lists.openshift.redhat.com http://lists.openshift.redhat.com/openshiftmm/listinfo/users
