Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

Hi Lida,

Q1. How will the compute node achieve HA when one instance fails in a compute host?

Controller are HA thanks to HAProxy in Mirantis because when one fails another node will be taken next. For compute node, you can also indeed create multiple nodes but that will not perform HA as you think. Of course, when you need to create a vm and one of your compute is down, another will be chosen by Nova Scheduler (so HA for vm creation). But in the case where you have already a vm running in your compute and the instance itself fails, you have to perform a feature by yourself. It exists many different manners to do it. A manner to proceed :

The most important things are to :

  • Monitor your VM, ping, list of the services running (depending on what service you consider to say it's a working instance).
  • Fence the VM and recover it. In Nova, I can only advice you to use the Evacuate feature which is used when you want to evacuate a vm from a failed host to another one.

Q2. I have understood the basics that Fuel master node is only used for deploying OpenStack. However if the fuel master node failed (e.g. the server got physically damaged), 1) will the OpenStack environment be damaged anyhow? If the fuel master node failed, nothing should happen. Fuel Master is indeed used for deployment. It should work normally if the fuel master is not up. If you can't access without the fuel master up, there is a misconfiguration. Frequently, because of the network, I know that Neutron is very not easy. for example..

2) will I be able to recreate the Fuel node easily as part of disaster recovery?

Fuel node is independant from your cluster. It will detect all your nodes available if your network is well configured. If you reinstall your fuel, there is no consequence (in theory).

Hi Lida,

Q1. How will the compute node achieve HA when one instance fails in a compute host?

Controller are HA thanks to HAProxy in Mirantis because when one fails another node will be taken next. For compute node, you can also indeed create multiple nodes but that will not perform HA as you think. Of course, when you need to create a vm and one of your compute is down, another will be chosen by Nova Scheduler (so HA for vm creation). But in the case where you have already a vm running in your compute and the instance itself fails, you have to perform a feature by yourself. It exists many different manners to do it. A manner to proceed :

The most important things are to :

  • Monitor your VM, ping, list of the services running (depending on what service you consider to say it's a working instance).
  • Fence the VM and recover it. In Nova, I can only advice you to use the Evacuate feature which is used when you want to evacuate a vm from a failed host to another one.

Q2. I have understood the basics that Fuel master node is only used for deploying OpenStack. However if the fuel master node failed (e.g. the server got physically damaged),

1) will the OpenStack environment be damaged anyhow?

If the fuel master node failed, nothing should happen. Fuel Master is indeed used for deployment. It should work normally if the fuel master is not up. If you can't access without the fuel master up, there is a misconfiguration. Frequently, because of the network, I know that Neutron is very not easy. for example..

2) will I be able to recreate the Fuel node easily as part of disaster recovery?

Fuel node is independant from your cluster. It will detect all your nodes available if your network is well configured. If you reinstall your fuel, there is no consequence (in theory).