search

viccross / ansible-playbooks

My collection of Ansible playbooks for various tasks (dominated by Openshift, currently)
Apache License 2.0
2 stars 3 forks source link

Setup for Infrastructure Node #31

Closed viccross closed 6 months ago

viccross commented 3 years ago

Post-install, the bootstrap node is no longer required and is a waste of resources. We were thinking for a time that the disk space could be added to the Bastion to allow a persistent storage location (using NFS) to be defined. A better use of the full resource capacity of the node would be to convert it to an Infrastructure node.

The virtual machine name can be changed from OCPZBOOT to OCPZINF0, which makes more sense for after the bootstrap node is finished with.

viccross commented 3 years ago

This requires a bit of discussion. It might be that a simple PoC cluster could get by with a single infrastructure node, but it would introduce a SPOF so it definitely would not suit for a larger or more productionised cluster.

Adding further nodes after install is not catered for by the process now. There is the additional resource that would be needed by these nodes (how to add DASD/CPUs/memory etc) as well as how to create configuration files etc. These and other points would be part of the whole "FastVM Productionisation" discussion.

viccross commented 3 years ago

The recommended guides for infra-node creation state that a number of nodes are needed, but they don't say that the ingress-router will not come up properly if less than two infra nodes are present. So, given that the two initial workers are symmetrical and the bootstrap node is the same config as the masters the infra provisioning could be handled by staging the conversion of bootstrap to worker and the existing workers to infra.

viccross commented 3 years ago

When node names are changed with #42 it would make sense for this to be included too (e.g. use WRK0 as the guest name for the bootstrap node).

viccross commented 6 months ago

Taking a different approach for this -- since three workers is a common requirement, I'm rebuilding the bootstrap into a third worker. The new refactored RHOCP workflow offers "Infra Node" deployment, which will create two infra nodes (initially as extra workers) for the cluster.