ESXI 6.5 unmap question

[XinXiaGeng]

 

hi, i am a software engineer。I am implementing iscsi's unmap interface.

This interface is used in vmware ESXI6.5.

When I tested the unmap function on esxi 6.5, I hang my own implemented iscsi lun to vmware. Format lun into VMFS 6.81 and set reclaim-priority=high. When I built two 10GB thick provisioning virtual machines, this virtual machine did take up 20GB of space in the backend storage. When I deleted the virtual machine, I configured a failure exception on my storage interface. At this time, vmware is also disconnected. The virtual machine on vmware has been deleted. The lun space on vmware also shows recycling. After clearing, After the vmware is reconnected, vmware will issue an unmap operation for part of the space when the vmware is reconnected. At this time, I deleted the 10GB virtual machine, but the backend only received an unmap operation of about 2GB of space.

Each time the switch is switched, it will receive an unmap size operation of approximately 500M to 2GB, but there are overlapping unmap intervals. After operating about 10 times, it was found that the unmap operation received in the backend storage did not cover 10 GB. Set the esxi iscsi datastore to high priority, shouldn't the vmware recycle all the space right away?

Or do i have to wait for a while? What is the recovery strategy for ESXI 6.5 when unmap is disconnected and reconnected?

 

[peetz]

Greetings and welcome to the forums!

According to the docs (Space Reclamation Requests from VMFS Datastores ) ESXi 6.5 does only support and use the "Low" space reclamation priority, even if you configure "Medium" or "High".

Have you also tested with ESXi 6.7?

Andreas

[XinXiaGeng]

hi，

I set the iscsi datastore to high priority in the ESXI background. In the web interface of ESXI 6.5, it only supports low priority, and the background can be set to high priority. If the default is used, it is really low priority. I remember that the low priority is to issue an unmap operation while waiting for the vsphere to be idle. The low priority guarantee is to issue all unmap operations in 0-24 hours. The specific document I forgot to save, I did find it in the vmware doc, said that the background setting to high priority is to immediately reclaim all the space.

The specific document I forgot to save, I am indeed I found it in the vmware doc. The above description says that setting the background to high priority is to reclaim all the space immediately. After testing this function, the high-priority, back-end storage can immediately receive the unmap command sent by ESXI 6.5 through the iscsi interface, and the low priority cannot be.Do you know that unmap will issue all unmap operation strategies in the case of 6.5 low priority or high priority? Is there any time strategy for issuing all unmaps?

[peetz]

Sorry,

I do not know the internals of the ESXi unmap code, so I cannot really answer your question. I suggest that you open a Support Request with VMware to get more information.

In the meantime the only useful resource I found is this blog post describing how you can monitor the unmap operatins in ESXi.

Andreas