VMware standard vSwitch port group configuration when ESXi host is added to CloudStack #3156
Description
ISSUE TYPE
- Improvement Request
COMPONENT NAME
- VMware vSphere / ESXi under CloudStack
CLOUDSTACK VERSION
- Any
- Tested in CloudStack 4.9 and 4.11
CONFIGURATION
- New ESXi host being added to existing cluster
OS / ENVIRONMENT
- Any
SUMMARY
- A newly added ESXi host added to an existing ESXi cluster utilising standard vSwitches will come online without any port group configuration.
- Port groups are added when a VR and / or VM is started on that host
- However if VMware native HA is utilised and an HA even occurs then a non-configured host can be chosen as failover destination without CloudStack being in the decision chain. At that point VMs will fail to fall over.
- The scenario is not unfeasible - it is not uncommon to either rebuild a host or introduce a new host as part of a a HA scenario. A repeat of the HA event will then lead to VMs being left offline.
Suggested solution:
- Upon ESXi host addition the existing cluster should be checked for port group configurations, and all required port groups configured on the new ESXi host.
STEPS TO REPRODUCE
- Configure a native HA VMware vSphere cluster in CloudStack.
- Put all hosts apart from one into maintenance mode.
- Now add a brand new, unconfigured host into the cluster.
- Power down the remaining ESXi host with running VMs out of band / pull power / ungracefully shut it down.
- Monitor the native VMware HA failover.
- VMs will fail to migrate to the newly configured host.
EXPECTED RESULTS
- Once added to a CloudStack managed cluster all ESXi hosts should be configured and ready to take on workloads, even if these are triggered out-of-band from CloudStack.
ACTUAL RESULTS
- HA failover to unconfigured host will fail.