fuel-library/deployment/puppet/vmware/spec/spec.opts
Igor Zinovik 47c5088be0 configure nova-compute use one vsphere cluster
- create separate directory '/etc/nova/nova-compute.d' where all
  configuration files that hold all configuration stanzas related to
  vmware compute driver
- in simple mode we upload init script in appropriate directory and
  create symbolic links on it that are named like
  nova-compute-vmware-%cluster_id% ; script internally splits its own
  file name and extracts cluster name.  This allow user to selectively
  start/stop nova-compute services.
- Ubuntu init system (Upstart) requires initd service to be restarted,
  on CentOS we run '/bin/true'
- create individual nova-compute configuration file per vSphere cluster;
  file is named `/etc/nova/nova-compute.d/vmware-%cluster_id%.conf',
  e.g.  /etc/nova/nova-compute.d/vmware-0.conf
- create appropriate corosync resource/primitive per nova-compute
  service in case of HA
- currently we use same login and password for all nova-computes
- set more strict access permissions (0600) on nova-compute.conf and
  generated files in `nova-compute.d' directory; these files hold
  sensitive information (credentials to vCenter server)
- set 'multi_host' configuration stanza to False for both deployment
  modes (nonHA/HA)
- add basic unit test coverage for provided functionality

DocImpact
Change-Id: I675dd2b7cdeacae1e703b82001dc2c855511e320
Impelements: blueprint 1-1-nova-compute-vsphere-cluster-mapping
2014-12-17 22:52:14 +03:00

2 lines
9 B
Plaintext