Abiquo Documentation Cookies Policy

Our Documentation website uses cookies to improve your experience. Please visit our Cookie Policy page for more information about cookies and how we use them.


Abiquo 4.7

Skip to end of metadata
Go to start of metadata

Introduction

Abiquo High Availability (HA) is a vendor-neutral automatic VM restart capability that allows enterprises deploying multi-hypervisor datacenters to have a cost-effective high availability mechanism. Abiquo High Availability is based on Abiquo's VM management features and is designed to complement hypervisor HA.

No vCenter Support

Abiquo High Availability is compatible with VMware vCenter. Choose either Abiquo HA or VMware High Availability with vCenter.

Compatibility

(error) Abiquo High Availability does not support XenServer
(error) Abiquo High Availability is compatible with vCenter. Choose either Abiquo HA or VMware High Availability with vCenter
(error) Abiquo HA will not failover a VM from an ESX 6 host to an ESX 5 host, or vice versa. It will not failover from an ESX 4.x to ESX 5 host, or from an ESX 5 to ESX 4.x host. Abiquo will failover between two hosts of the same major version and in order to use Abiquo HA, all the hosts on the same rack should be of the same major version.

The basic functionality is that when a hypervisor failure is detected in a rack with HA enabled, the VMs created in Abiquo and running in that hypervisor from a shared datastore are restarted in another host in the same rack. High Availability will function with external storage and persistent VMs.

Although some of the HA configuration is performed within the Abiquo GUI, the system administrator should understand the overall configuration of the HA system as described in the outline below.

HA Prerequisites and Configuration

  • HA-Enabled Rack
  • Physical machines with IPMI Configured (Except on Managed Racks)
  • Multiple Hypervisors of the Same Type (Except XenServer)
  • VMs Created in Abiquo (Not Imported)
  • VMs Deployed from a Shared Datastore
HA-Enabled Rack

Enable High Availability in the Abiquo GUI when Creating Racks.

Physical machines with IPMI Configured

Physical machines on a standard rack enabled for HA must have IPMI configured. Physical machines on managed racks do not require IPMI because they will be controlled using the managed rack API (UCS racks only).

Configure IPMI in the Abiquo GUI when Creating Physical Machines. You will need to enter the IP address, Port (optional), User and Password.

Multiple Hypervisors of the Same Type

Abiquo supports HA on standard hypervisors except XenServer because of restricted datastore access. See the corresponding Private cloud providers documentation for further details.

In order for HA to function in a rack, there must be more than one hypervisor of the same type in the rack.

VMs Created in Abiquo or Captured

Abiquo supports High Availability of VMs created and deployed in Abiquo. It also supports High Availability of captured VMs. Abiquo High Availability will not function with imported VMs that have not been captured.

VMs Deployed from a Shared Datastore

The VMs must be deployed but they can be in any of the deployment sub-states (on, off, paused or rebooted). The HA hypervisors must all have access to a shared datastore to allow them to restart the same VMs. The hypervisors may mount the same datastore with different names and using different technologies (e.g. NFS or CIFS). Abiquo uses a UUID folder to identify the datastore and this folder must not be deleted. If an HA hypervisor has more than one shared datastore, Abiquo will deploy using the standard datastore scheduler. Abiquo HA will restart VMs that are deployed on the shared datastore.

Enable the same shared datastore on each hypervisor on the HA Rack in the Abiquo client when Creating Physical Machines.

Detecting Host Failure

Abiquo uses ping to periodically check the hypervisors in the physical machines in the Rack. If a hypervisor does not respond to one check, it is marked. If it does not respond to the following check, the HA process starts.

Configuring the Check Frequency

The frequency with which the system should check the status of the hypervisor hosts is configurable in abiquo.properties file (See Abiquo Configuration properties).

  • The Infrastructure Check runs to check all physical machines 
    • How often it checks is configured with the abiquo.nodecollector.infcheck remote services property (default value is 5 minutes)
    • By default the Infrastructure Check runs once every 5 minutes, which is recommended to prevent HA being triggered for temporary outages such as a machine rebooting. This check should always be set to a value above 1 minute in order to avoid network saturation.
    • The Infrastructure Check is part of the Discovery Manager (nodecollector) remote service
  • If a machine fails a number of checks, HA is triggered and the machine state is set to state HA_IN_PROGRESS. 
    • Failed checks configured with the abiquo.ha.infcheck.cycles property (default value is 2).
  • The Abiquo HA Check runs to check if the VM tasks have finished on physical machines in state HA_IN_PROGRESS
    • If the tasks have finished, the machine is set to DISABLED_BY_HA
    • How often it checks is configured with the abiquo.ha.check.delay Abiquo server property (default value is 300000 milliseconds, which is 5 minutes)
    • The Abiquo HA Check is part of the Abiquo Server

The High Availability Process

When High Availability is enabled on a Rack, Abiquo periodically checks the hypervisors in the physical machines of the Rack (default 5 minutes).
If a hypervisor does not respond after the number of checks configured (default 2), Abiquo:

  • Shuts down the physical machine using IPMI
  • Makes a shortlist of suitable hosts that meet HA prerequisites
  • Restarts the VMs on these hosts using the same shared datastore
  • Marks the failed host as DISABLED_BY_HA

If there are no suitable hypervisors available (e.g. the whole Rack is affected by the system failure), the HA process will not run. If there are not enough resources on the available hypervisors, the VMS will not be restarted on a new hypervisor and will remain unchanged on the failed hypervisor.

See How Abiquo HA Works

Re-enabling Physical Machines Disabled by High Availability

You must manually restart a hypervisor that failed and was shut down by Abiquo using IPMI or the managed rack API.

Attention: Restarting an ESX or ESXi host disabled by HA

Before manually enabling an ESXi host in Abiquo, first check the hypervisor configuration using vSphere Client to ensure that the datastores are active. Perform a rescan if necessary. Then check for VMs with no name that are in the UNKNOWN (INVALID) state. Delete these VMs in the hypervisor. Do not delete VMs with proper names because the platform will manage these VMs.

The hypervisor will have a state of DISABLED_BY_HA in the Abiquo GUI. The Cloud Administrator must then start up, perform checks and manually enable the physical machine in Abiquo. This machine will then be shown as MANAGED by Abiquo. Abiquo will reactivate any VMs previously deployed on it that could not be moved. The machine will also be available for new deployments.