Unmount datastores - file system is busy

[andvm]

Hi,

Am in the process of unmounting a number of datastores from vcsa 6.0 (all VMs migrated from them) but this is failing for a small number of hosts due to "File system is busy" message.

I am going to follow the steps shown in this link VMware Knowledge Base but not sure what is the impact on the hosts if the dumpfiles are removed? (Will it automatically create a new one on a new random datastore next time)

How can I check if the datastore is used as an HA heartbeat as when I checked the HA configuration the "Use datastores from the specified list and complement automatically if needed" but none of the datastores is ticked. Would this prevent me from un-mounting the datastore or would HA pick up another datastore automatically without any impact?

The datastore is not part of a Datastore Cluster and Storage I/O Control is disabled for the datastore, let me know if you have other suggestions as suspect that there are files on the datastore that are still actively used/locked by one or more ESXi hosts. Want to avoid rebooting the hosts which cannot have the datastore unmounted from.

Thanks

[a_p_]

... but none of the datastores is ticked

HA picks two shared datastores for heartbeating. To ensure non of the datastores that you want to unmount, simply pick two other datastores.

Also make sure that none of the VMs points to the datastores with e.g. an ISO image that's configured for their CD-ROM drive. Even a previously attached ISO image that's currently not connected, but has been connected in an existing snapshot will prevent from unmounting a datastore.

André

[andvm]

The issue seems to be caused by vsantraces as can see open files pointing to the datastore UUID from the hosts on which the datastore unmount failed.

I found this document online that should help VMware Knowledge Base so I will try to below command from the same hosts and specifying a different datastore name or UUID unless you are aware that this causes any downtime or issues to the vSAN environment...

esxcli vsan trace set -p datastore_name

It might be that each host will automatically select an available datastore to store vsantraces but hopefully by running the above command it will close any open vsantraces file it has and create new ones on the new datastore. This should let me unmount the datastore.

If successful should the next steps be to detach the datastores and then rescan each and every single host that had access to the same datastores before these are unpresented?

Thanks

[andvm]

Update: the above worked for most of the hosts except one (which probably I will end up restarting). The strange thing is that when listing the datastore contents I cannot see any files that are recent or in use so entirely not sure why it fails to unmount from the specific host. No ISO file exists on the datastore.