Incorrect Output from HVM Guest Sharing Multipath-based Disks
(Doc ID 1120616.1)
Last updated on FEBRUARY 27, 2019
Applies to:Oracle VM - Version 2.2 and later
Oracle Cloud Infrastructure - Version N/A and later
Multiple "Hardware-Virtualised" (HVM) guest nodes were setup to share multipath devices as documented in the VM Manager User Guide. Writes to the devices are not seen by HVM guests that run on other VM Servers, and are not seen by any "Para-virtualised" (PVM) or HVM guests that utilise "Para-virtualised" device-driver modules (PVHVM), regardless of whether the PV/PVHVM guests are running on the same VM server. PV and PVHVM guests operate correctly among themselves, but are affected if sharing a device with an HVM guest.
The problem came to light when the devices were used for OCFS2 volumes.
The OCFS2 disk heartbeat writes from one HVM node are not seen by another node.
This causes OCFS2 errors, including file corruption as nodes do not recognise each other as have a volume mounted. Possible symptoms are messages such as:
(o2net,1741,0):o2net_connect_expired:1664 ERROR: no connection established with node 6 after 30.0 seconds, giving up and returning errors.
One or more of the nodes in the cluster might "self-fence" (reboot themselves).
Implemented "Multipath-based Disks" shared to multiple guests.
The problem does not affect "File-based Disks".
To view full details, sign in with your My Oracle Support account.
Don't have a My Oracle Support account? Click to get started!