Unable to take snapshot of VIC 1.4.1 which was upgraded from 1.3.1, as the disks which got moved are marked as independent

[lgayatri]

@zjs @hickeng @renmaosheng

Steps:

1. VIC OVA 1.4.1 was successfully upgraded from 1.3.1 few weeks ago
2. Attempted to take a snapshot on 1.4.1 and the operation failed due to Hard Disk 2, 3 and 4 being marked as independent disks.
   

Upgrade from 1.3.1 to 1.4.1 was done using upgrade.sh and following steps https://vmware.github.io/vic-product/assets/files/html/1.4/vic_vsphere_admin/upgrade_appliance_manual.html
(disks were not manually moved)
upgrade.log

[zjs]

@andrewtchin, was this considered as a part of the design for the new automated upgrade process? A quick search of GitHub didn't show anything, but I wanted to confirm I'm not missing something obvious before digging into this more deeply.

[andrewtchin]

This was not considered in upgrade design

[wjun]

The strange thing is the upgraded disk with 'independent' mode is read only and cannot be changed to other modes in powered-off state.

[zjs]

It's not clear why the disk mode changes to independent. KB 1007532 provides instructions for changing the mode back to dependent, which requires that the VM is powered off. However, for some people the UI still shows the mode as read-only (e.g., as noted in #1885 (comment)). A work-around for this seems to be to use the VIM API to initiate the change instead of the vCenter UI. This can be achieved using govc vm.disk.change -mode persistent.

[zjs]

It seems like the the govc vm.disk.attach command used to attach the copied disks to the new OVA infers the disk mode from the supplied -link (default: true) and -persist (default: true) arguments.

https://github.com/vmware/govmomi/blob/8d973c31513ae5f16bdcf5d88c1b8ae2267bc63d/govc/vm/disk/attach.go#L105-L120

When we call govc vm.disk.attach, we provide neither argument:

vic-product/installer/build/scripts/upgrade/upgrade.sh

Lines 391 to 395 in e811e0a

	govc vm.disk.attach -vm="$NEW_VM_NAME" -ds "$NEW_DATASTORE" -disk "$NEW_DATA_DISK" || (log "Failed to attach data disk" && exit 1)
	if [ "$ver" != "$VER_1_2_1" ]; then
	govc vm.disk.attach -vm="$NEW_VM_NAME" -ds "$NEW_DATASTORE" -disk "$NEW_DB_DISK" || (log "Failed to attach database disk" && exit 1)
	govc vm.disk.attach -vm="$NEW_VM_NAME" -ds "$NEW_DATASTORE" -disk "$NEW_LOG_DISK" || (log "Failed to attach log disk" && exit 1)
	fi

Perhaps supplying -link=false will fix this.

[zjs]

Adding kind/note as it seems like we'll need a release note (or perhaps KB?) to provide a work-around to users of 1.4.0 and 1.4.1.