3.6 Cost-effectiveness (HA - High Availability)

The price of the product(High Availability, HA) Silver is a function that allows a system or service to operate continuously and without interruption even in a situation of disruption. VirtOn The Proxmox It manages the availability of nodes and resources within the cluster based on HA structures.

VirtOn The HA page provides the following features.

• HA monitoring of the condition
• List of HA resources
• HA Advanced setup (Affinity Rules)

Before use

3.6.1 HA monitoring of the condition

In the HA state monitoring area, each node in the cluster is Check the HA role, service type, status in real timeWe can do that. VirtOn The Proxmox Based on HA information, CRM(Cluster Resource Manager) And LRM(Local Resource Manager) And then we're going to tell you what's going on.

3.6.1.1 CRM (Master node.) It's in the air.

CRM is a cluster of controlled HA resources. Central control nodeAnd in the cluster, we have There's only one CRM master node.It does.

Display items

• The node. The Master Node is currently performing the CRM role
• Role : CRM.
• Service : Control
• It's in the air. (Quorum is in effect.)
  • Quorum top (OK)
    • Cluster polynomial is in normal state.
    • It's green.
  • Failure (No Quorum)
    • Quorum is lost.
    • It's red.
    • HA is uncontrolled.

3.6.1.2 LRM (The HA state of the node.)

LRM(Local Resource Manager) Silver is the component that actually manages the HA resources at each node. Included in the cluster All nodes are displayed in LRM stateAnd then the node is Name by name(A–Z) So it's sorted into.

Display items

• The node. Name of node that comprises HA
• Role : LRM
• Service : HA mode of operation of the node
• It's in the air. : Node-specific HA state

3.6.1.3 LRM state type and meaning

⚠️ Beware of it. Lost Lock It's not just a state of rest. HA operation is over.And that's what we're doing. We need to check for resource issues or network/agent issues.

3.6.1.4 Data loading and exception processing

• HA status information is Proxmox API It's viewed in real time through the Internet.
• During the state loading, Loading spinner.It's going to be shown.
• If an error occurs during the HA status check:
  • An error message is displayed.
  • [Try again] buttonYou can re-evaluate it immediately.
• If the HA node is not available:
  • There's no nodes that are registered. The guide phrase is displayed.

3.6.2 HA Resources list

The HA Resource List is a display of the value of the resource.(HA) This is the application of VM And you can check and manage container resources at a glance. Each resource is automatically recovered and re-allocated when it fails.

3.6.2.1 Column description

3.6.2.2 Resource type (Type)

HA resources have the following two types.

3.6.2.3 Resource state (Type)

HA resources are visualized in color and text depending on the current state.

• When creating HA resources, it is used in four states. (Default started)
  • Running state (Started) : Generate a normal running HA resource
  • Stop it. (Stopped) : generate the HA resource in the off state
  • Ignored (Ignored) : HA resource generation temporarily ignored by the policy
  • Deactivation (Disabled) : HA resource generation excluded from the management of HA

3.6.2.4 Maximum restart (Max Restart)

When there's an error or an abnormal shutdown in the process of running a resource, Maximum number of times you try to run it again automatically at the same nodeThat's it.

Example of action

• Maximum restart is equal to 1
• VM This extraordinary end
• HA's trying to restart the first time at the same node.
• The policy after the reboot failed(Re-allocated) Keep moving.

If the number of reboots exceeds the limit, the re-allocation or Fence policy applies.

3.6.2.5 Maximum re-allocation (Max Relocate)

When a restart failure or node failure occurs, Maximum number of times a resource can be moved to another nodeThat's it.

Example of action

• Maximum re-allocation is equal to 1
• Current node disruption occurrence
• To another node. VM Move it around.
• Additional relocation stop when over-allocation

It's a protective device to prevent excessive movement in a cluster environment.

3.6.2.6 Failback

After the node that was damaged was restored, The option to set whether to return the resource to the original run nodeThat's it.

• On: The role of the node in a clear environment (The workload.)
• OffStability is a vital service (Preventing unnecessary movement)

3.6.2.7 Resource selection and management

• I'm on the list. Only one resource is available.
• The next option is to work.
  • Correction Add and Module items are the same, single ID changes are prohibited
  • Remove it : Remove the window output
• If you don't select it, the control button will be disabled.

3.6.2.8 HA Resource input entries and permissible range

3.6.3 HA Advanced setup (Set up the Affinity Rules)

In the HA advanced settings, Affinity Rules(The rules of friendship) Through You can tightly control the execution location and deployment policy of HA resources.

Affinity rules are basically two things.

• The rules of node friendship (Node Affinity Rules)
• The rules of resource-friendliness (Resource Affinity Rules)

Each rule directly affects the deployment stability and disruption response strategy of HA resources.

3.6.3.1 Node-friendly rules (Node Affinity Rules)

The rule of nodal friendliness is And what kind of node is the priority to run a particular resource?I'm going to define it. We can use weighing and policy to set the priorities for implementation and enforcement.

3.6.3.1.1 Targeted resource

• The rules are: What HA resource is that?(VM/CT) It applies toIt's a representation of the
• You can specify one or more resources, and you can use the comma.(,) And then we're going to divide it into:
• Included in the target resource VM/CT And that's the rule.

Operating point

• Core service VM It's recommended to specify a clear rule of node friendliness.
• Testing or critical resources can minimize the application of the rules.

3.6.3.1.2 List of nodes (Weight)

The list of nodes is The node that the resource can run and its prioritiesI'm going to define it.

The concept of weight

• We're going to show that weighting is a number.
• The higher the value, the more the first run object.So this is it.
• If the weight is the same, it's sorted by node name.
• The weighting 0 And even if it is, it's explicitly shown.

An example of how to behave

• phum01 → Priority run node
• phum03 → The chain run node
• phum02 → The last candidate node

When a fault occurs, HA tries to run from a high-weight node.

3.6.3.1.3 Policy (Strict)

The policy is to follow the rules of node friendliness. Whether or not to force it.It determines the

It's strict. (Strict)

• It's not running except at the specified node.
• If all the specified nodes are not possible, the resource will not run.

Recommended use scenarios

• There are restrictions on licensing. VM
• Specific hardware-dependent services
• Fixed role node configuration

Preference (Preferred)

• First try running the node based on the
• If all nodes are unstable, run automatically to another node.

Recommended use scenarios

• General services VM
• The environment that gives priority to value-added services

3.6.3.2 Rules of resource friendlyness (Resource Affinity Rules)

The rules of resource-friendliness are: The execution relationship between resourcesI'm going to define it. You can run specific resources together, or you can run them separately.

3.6.3.2.1 Objective resource

• The rules apply to the Several HA resource groupsIt means.
• The designated resources operate as a single policy unit.

Policy 3.6.3.2.2 (Affinity)

Run along. (Positive)

• It's a service configuration that needs to minimize network latency.
• Example: Web server + cache server

Separate run (Negative)

• Configuration to reduce the impact of disabilities
• Examples: Active/Standby, dual service

How to solve problems