ESXi host part of vSAN cluster taking long time to boot.
Host some time freezes and esxcli
commands execution would timeout or fail.
VMware vSAN 7.x
VMware vSAN 8.x
The issue is seen when there are VMs registered on ESXi host from vSAN datastore and the underlying objects either not available or inaccessible.
When the namespace or VMDK objects are in unavailable state, since the VM is registered on host, it keeps querying the object which is either unavailable on vSAN or inaccessible. This makes the hostd experience issues and cause the host operations to fail or take time.
When checking the vmkernel.log on ESXi host, the below events would show up.
[root@esxi:/var/log] less vmkernel.log
2025-04-17T01:43:21.642Z cpu33:2104783 opID=888aa9af)Vol3: 1005: Unable to register file system bd342164-9089-9070-3280-######## for quesce timeout notifications: Inappropriate ioctl for device
2025-04-17T01:43:21.675Z cpu33:2104783 opID=888aa9af)Unmap6: 7170: [Unmap] 'bd342164-9089-9070-3280-########': could not obtain partitionInfo for device 0x43086adaacf0: Inappropriate ioctl for device
To fix the issue the below steps can be used.
Example object here is: a646fc67-407f-0c72-ce20-########
localcli vsan debug object list -u a646fc67-407f-0c72-ce20-########
This may not report any output.
cmmds-tool find -t DOM_OBJECT -u a646fc67-407f-0c72-ce20-########
This may show blank report.
/etc/vmware/hostd/vmInventory.xml
, the VM is registered and have the object which hostd is querying on a continuous basis.[root@esxi:/var/log] cat /etc/vmware/hostd/vmInventory.xml
<ConfigRoot>
<ConfigEntry id="0011">
<objID>55</objID>
<vmxCfgPath>/vmfs/volumes/vsan:52844907ce602496-########/bd342164-9089-9070-3280-########/vm.vmx</vmxCfgPath>
</ConfigEntry>
</ConfigRoot>
[root@esxi:/var/log]
If there is a STALE VM which has been deleted from vSAN, however the entry stays in the host configuration (vmInventory.xml), please unregister the VM from the host.
Unregister the VM from the host and make sure vmInventory.xml does not have the STALE entry to resolve this issue.