In a VMware vSAN stretched cluster running vSphere 7.0 U1 or U2, high latency or host disconnects are experienced when attempting to interact with virtual machines within the vSAN cluster

Products

• VMware vSAN

Issue/Introduction

Symptoms:

• vSAN 7.0 U1/U2 stretch cluster reports objects as non-compliant with no rebuild.
• The EPD service may crash and won't stay running after restart.
• The vmkernel.log may contain entries like the following:
  • Rejoin to Discovery: (Reason: Arena memory exhausted)

Environment

VMware vSAN 7.0.x

Cause

• Objects greater than 255GB created while the witness is unavailable may span both sides of the stretched cluster and remain unprotected by the selected host failures to tolerate (HFT) or site failures to tolerate (SFT) specified in their storage policy.
• Once the witness becomes available, vSAN will attempt to bring objects into compliance with their specified storage policy.

Resolution

• Upgrade to vSAN 7.0 U3c+ at the earliest opportunity.
• If any VM guests are impacted, please engage VMware support for assistance.

Additional Information

Impact/Risks:

• Loss of availability to data
• Hosts become non-responsive due to CMMDS arena memory exhaustion
  • Hosts only become responsive after a reboot.