Ceph » Orchestrator

https://access.redhat.com/webassets/avalon/d/Red_Hat_OpenStack_Platform-13-Spine_Leaf_Networking-en-US/images/0bd8baa06bb1d0b101171eddf3df1582/OpenStack_Spine_Leaf_466050_0218_routed.png

Problem: different daemons of e.g. RGW need to bind to different networks

• https://github.com/ceph/ceph-ansible/blob/master/roles/ceph-facts/tasks/set_radosgw_address.yml
• https://github.com/ceph/ceph-ansible/blob/master/roles/ceph-config/templates/ceph.conf.j2
• example workarounds we're doing in ansible https://review.opendev.org/c/openstack/tripleo-ansible/+/773364

having a "mapping" between hosts and subnet (instead of a str) would help for spine and leaf deployments:

• host1: storage_subnet_1
• host2: storage_subnet_2
• ..
• hostN: storage_subnet_N

workaround at the moment:
1. apply the cluster spec
2. call a bunch of ceph config set ... commands

• https://github.com/ceph/ceph/pull/39648

problem: grafana, ganesha (only a config file for all the daemons)

• grafana: https://community.grafana.com/t/grafana-wont-bind-to-an-ip-address/6063/6
• again: need to pick an ip from a subnet list
• ganesha: https://github.com/ceph/ceph-ansible/blob/246e31c0d3c3dd16cdcf2a1e6d54e85c857ff8bd/roles/ceph-nfs/templates/ganesha.conf.j2#L8-L12
• some components like prometheus/alertmanager apply the ip address as cli argument (set in the systemd unit)
• https://github.com/ceph/ceph-ansible/blob/master/roles/ceph-prometheus/templates/prometheus.service.j2

dashboard

• https://docs.ceph.com/en/latest/mgr/dashboard/#host-name-and-port
• https://review.opendev.org/c/openstack/tripleo-ansible/+/773364/31/tripleo_ansible/roles/tripleo_cephadm/tasks/dashboard/grafana.yaml