VMware vSphere 6.5 – HA and DRS Improvements

There is some more information about other vSphere 6.5 enhancements as below;

vSphere HA Admission Control

• Simplified configuration workflow
• Choose Failure To Tolerate (FTT) and we do the rest

1. Based on % of resources reserved admission control policy
2. Automatic calculations
3. Overrides possible

• Issue warning if performance degradation could result
• Previously, HA would restart VMs, but could have caused performance impact without prior warning.

vSphere HA Restart Priorities

• VM Restart priorities determine the order resources are allocated in the cluster post-failure.
• Additional Restart Priorities added

1. Highest
2. High
3. Medium
4. Low
5. Lowest

• Additional priorities provide greater control to restart order
• Admission Control can prevent lower priority VMs from restarting

vSphere HA Orchestrated Restart

• Enforce VM to VM dependency chains

1. Great for multi-tier applications that require VMs to restart in a particular order
2. Improved application recoverability
3. Database < App < Web

• Validation checks

1. Detects circular dependency rules
2. Within and outside priority group

Proactive HA

• Hardware OEM participation required
• Detect hardware conditions of host components by receiving alerts from hardware vendor monitoring solution

1. Dell Openmanage
2. HP Insight Manager
3. Cisco UCS Manager

• Notifications of host impacted, its current state, error causes, severity and physical remediation steps
• vMotion VMs from partially degraded hosts
  Configurable based on type of failure

Quarantine mode

• In server health-aware vSphere clusters
• Quarantine mode is a new vSphere host state

1. Any degraded host will be placed in quarantine mode

• vSphere DRS will Evacuate virtual machines if

1. No performance impact on any virtual machine in the cluster
2. None of the business rules is disregarded

• Avoid placing virtual machines
• Only use this server if the virtual machine demand or any of the business rules cannot otherwise be satisfied

DRS Policies

• Even Distribution (Peanut Butter)
• Advanced settings exposed in the UI for simpler configuration

1. Improved availability by avoiding too many eggs in one basket

• Consumed Memory vs Active

1. Default: DRS uses Active memory + 25% for load balancing
2. Policy Active: changes the DRS advanced PercentIdleVMInMemDemand=100

• CPU over-commitment

1. Prevent any single host from becoming over-committed on CPU
2. Great for VDI environments

Network Aware DRS

• Adds network bandwidth considerations by calculating host network saturation (Tx & Rx of connected physical uplinks)
• Avoids a over-subscribing a host network links, although not guaranteed.  Best effort approach. CPU & MEM performance is still priorities over network.

SIOC + SPBM

• Managed using a policy via Storage Policy Based Management (SPBM)
• Storage IO limits are enforced using IO Filters (VAIO)

Categories: VMware

Victor Wu

Chief Architect, Blogger, Author at Dell EMC Knowledge Sharing & Packt