Checking the services
HAGISTACK’s own status (hagistack status)
Section titled “HAGISTACK’s own status (hagistack status)”status is a HAGISTACK subcommand. It is not the OpenStack CLI (the openstack command); it reports the state of a host HAGISTACK deployed.
The hagistack script is not on the PATH: run it with ./ from the directory you cloned (hagistack/rocky10.2). It needs root.
cd hagistack/rocky10.2sudo ./hagistack statusIt shows:
- the HAGISTACK version, and the target release and OS
- the role (
all-in-one,controller,compute), the RPM source and set identity - the SELinux mode now and at boot, and what HAGISTACK did to it
- the completed phases
- the state of every service unit (
activeand so on)
systemd
Section titled “systemd”systemctl status openstack-nova-computejournalctl -u openstack-nova-compute -n 100The main units:
| Unit | Node |
|---|---|
mariadb, rabbitmq-server, memcached |
controller |
httpd (Keystone, Placement, Neutron API, Nova API, Horizon) |
controller |
openstack-glance-api |
controller |
ovn-northd |
controller |
neutron-rpc-server, neutron-periodic-workers, neutron-ovn-maintenance-worker |
controller |
openstack-nova-conductor, openstack-nova-scheduler |
controller |
ovn-controller, openvswitch |
every node |
neutron-ovn-metadata-agent |
every node |
openstack-nova-compute, libvirtd |
every node with a hypervisor |
The OpenStack CLI
Section titled “The OpenStack CLI”sudo -i. /etc/hagistack/admin-openrcopenstack compute service listopenstack hypervisor listopenstack network agent listcompute service list: everythingenabled/uphypervisor list: every compute hostupnetwork agent list: every agentAlive=True
Re-runs
Section titled “Re-runs”all-in-one and compute-add are designed to be run again.
- Secrets: reused, never rotated.
- Databases: never dropped.
- Configuration: edited key by key; a service restarts only when its configuration is newer than its running process.
- Initial resources: show-or-create by name.
HAGISTACK does not roll back. Fix the cause of a failure and run the same command again. A re-run is not a reconfiguration tool: existing resources and endpoints are left as they are.