APD condition on some RDM LUNs

[xbradshr]

My co-worker has migrated some VMs off our ESX 4.0 hosts, to newer versions of vmware. These VMs had RDMs on the hosts. The RDM LUNS were unpresented from the hosts, without using LUN masking procedures. This leaves the RDM LUNs in an APD / all paths down state. This has not caused issues yet, but I am aware that APD can cause hosts to have problems with the hostd service., such as disconnects from vmware, VMs locking up etc.  I was  wondering the best way to correct this. I am aware that rebooting the host can fix it. My question is, is there any other way to remove the RDM LUNs that are now in an APD state?

[hussainbte]

I am sure you know that your vSphere is End of support.

VMware KB: End of Availability and End of Support Life for ESX 4.x and VMware Management Assistant 1...‌

Regarding the APD condition on ESX4 hosts, make sure the /vmfs/failopenifAPD is enabled to prevent other VMs having an impact(KB linked).

Unregister any unresponsive/inaccessible VMs and perform a rescan on host(we are expecting to clear up the dead paths hear)

VMware KB:    Virtual machines stop responding when any LUN on the host is in an all-paths-down (APD...

esxcfg-rescan -A

Please note if the host is a part of cluster, please migrate the VMs which are running fine to a different host to avoid any impact.

A host rescan in this situation can trigger a hostd issue if it is dormant now.

[kermic · Accepted solution]

I guess you are asking about proper cleanup after LUN is "unpresented" from host on either fabric or the array.

The host reboot should definitely clean this up.

If you can't reboot then check this post from Cormac Hogan: Best Practice: How to correctly remove a LUN from an ESX host - VMware vSphere Blog - VMware Blogs

It should give you an idea and some hints on how to clean up properly.

Hope this helps.