Note: This bug is displayed in read-only format because the product is no longer active in Red Hat Bugzilla.

Bug 1553154

Summary: QEMU image locking on NFSv3 prevents VMs from getting restarted on different hosts upon an host crash, seen on RHEL 7.5
Product: Red Hat OpenStack Reporter: Lee Yarwood <lyarwood>
Component: openstack-novaAssignee: OSP DFG:Compute <osp-dfg-compute>
Status: CLOSED NOTABUG QA Contact: OSP DFG:Compute <osp-dfg-compute>
Severity: high Docs Contact:
Priority: high    
Version: 14.0 (Rocky)CC: ahino, akrejcir, aliang, alukiano, amureini, areis, berrange, chayang, coli, cshao, dasmith, eglynn, famz, fromani, jhakimra, juzhang, kchamart, knoel, kwolf, lsurette, mbooth, michal.skrivanek, michen, mkalinin, msivak, mtessun, ngu, nsoffer, owalsh, pingl, qzhang, rbalakri, rjones, sbauza, sferdjao, sgordon, srevivo, stirabos, virt-maint, vromanso, xuwei, ycui, yhong, ykaul, ylavi, yzhao
Target Milestone: ---   
Target Release: ---   
Hardware: x86_64   
OS: Linux   
Whiteboard:
Fixed In Version: Doc Type: If docs needed, set a value
Doc Text:
Story Points: ---
Clone Of: 1547095
: 1556957 (view as bug list) Environment:
Last Closed: 2018-03-28 23:54:50 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: Virt RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:
Bug Depends On: 1547095    
Bug Blocks: 1556957    

Comment 1 Nir Soffer 2018-03-08 12:35:43 UTC 
In RHV this was fixed by using the "nolock" mount option for NFS v3 or
autonegotiate. Please see
https://github.com/oVirt/vdsm/commit/8aad4ebad10bab15c46ff743c105b6c2cba68d4e
Comment 2 Matthew Booth 2018-03-08 12:52:26 UTC 
Note that we can also deploy Nova's instances directory on NFS, so it's not just a Cinder thing. Nova takes explicit filesystem locks in here extensively. 'nolock' would make all of these locks host-local, which for the most part is likely all they were intended for anyway. However, in the image cache these are likely intended to be multi-server, and I would not be surprised to discover other cases.

TL;DR We cannot assume that setting nolock is a safe thing to do for Nova ephemeral storage. It might be, but we'd have to carefully audit all use of local filesystem locks.