Closed puppetninja closed 5 years ago
Hello, @PuppetNinja! Could you please describe your use case? What is the purpose of shutting down the VMs? Would you also need the ability to spin them back up again?
It turns out we may have something like this already for VSD and ES. During the upgrade procedure on KVM and vCenter hypervisors, we gracefully shut down the old VSDs and ES nodes without fully undefining or destroying them. To support this, we have a variable called "preserve_vm". You can get the behavior you desire by setting an extra variable on the command line. For example:
./metroae vsd_destroy -e preserve_vm=true
./metroae vstat_destroy -e preserve_vm=true
We do not have the ability to spin them back up. We do not have the ability to gracefully shut down VSC or NSGv.
Hi @bacastelli,
I am working on bringing up the Nuage lab in Westford, and I hit the issue for our lab VSD cluster, there was a planned downtime for power maintenance so we need to shutdown the VSD cluster, I shutdown one of the VSD node without a graceful shutdown first using monit as I was on a terminal connected to a wrong node.
And the VSD cluster failed to be back online again after we powered up the rack.
I am not sure whether this is a VSD product problem or Ops problem. But I think it would be less error prone to make that ops work automated.
I will try the solution you provided. Thanks !
In our testing, we shutdown all of the services before shutting down the VMs: monit -g vsd-stats stop monit -g vsd-core stop monit -g vsd-common stop
On each of the VSDs.
Then to bring them back up, we start them: monit -g vsd-stats start monit -g vsd-core start monit -g vsd-common start
and then on the primary VSD only, we set the master persona: /opt/vsd/sysmon/bootPercona.py --force
There is a "graceful shutdown" procedure somewhere in the Nuage documentation which explains this.
Hi @mpiecuch-nuage yes, I was wondering if the automation of the above followed by VSDs shutdown could be provided by a playbook in metro. In my case I forgot the above steps on one of the VSD node in a cluster and the cluster didn't come up after we power up the hypervisors.
And I do see the this playbook https://github.com/nuagenetworks/nuage-metro/blob/master/src/playbooks/with_build/vsd_services_stop.yml
For shutting down the services only, but it is already quite useful, thanks, I can close this one
Hi Metro Team,
Can we add support to gracefully shutdown some of the deployed Nuage components ? Especially for components like VSD clusters
Thanks