Bug 810082

Summary: [RFE] support hard mounts for NFS in addition to soft mounts
Product: [Retired] oVirt Reporter: Ayal Baron <abaron>
Component: vdsmAssignee: Itamar Heim <iheim>
Status: CLOSED WONTFIX QA Contact: Aharon Canan <acanan>
Severity: unspecified Docs Contact:
Priority: unspecified    
Version: unspecifiedCC: acathrow, amureini, bazulay, bugs, iheim, jkt, mgoldboi, nsoffer, tburke
Target Milestone: ---Keywords: Improvement
Target Release: 3.3.4   
Hardware: Unspecified   
OS: Unspecified   
Whiteboard: storage
Fixed In Version: Doc Type: Enhancement
Doc Text:
Story Points: ---
Clone Of: Environment:
Last Closed: 2014-08-04 09:25:19 UTC Type: Bug
Regression: --- Mount Type: ---
Documentation: --- CRM:
Verified Versions: Category: ---
oVirt Team: --- RHEL 7.3 requirements from Atomic Host:
Cloudforms Team: --- Target Upstream Version:
Embargoed:

Description Ayal Baron 2012-04-05 05:53:30 UTC
Description of problem:
Currently NFS storage domains are mounted with soft mount.  Although there is no technical issue with it, hard mounts are much more common hence tested better so it's preferable to support hard mounts.
To prevent vdsm from hanging support should be for hard,intr and have mechanisms to interrupt I/O in the control path after a timeout

Comment 1 Ayal Baron 2012-04-05 05:54:32 UTC
*** Bug 574733 has been marked as a duplicate of this bug. ***

Comment 2 Itamar Heim 2013-12-07 19:40:54 UTC
is this still relevant with all the time we are with soft mounts?

Comment 3 Ayal Baron 2013-12-08 10:47:02 UTC
(In reply to Itamar Heim from comment #2)
> is this still relevant with all the time we are with soft mounts?

It will always be relevant regardless of the current way being 100% ok.
1. With NFS people prefer their VM not pausing on EIO but hanging the I/O.
2. Perception wise people prefer hard mounts

On the flip side, changing everything to work with hard mounts might be challenging due to the need to detect hangs and interrupt processes such as dd (in the control path)

Comment 6 Itamar Heim 2014-08-04 09:25:19 UTC
don't remember anyone asking about this for a while. its making things complex for the spm/sdm data operations. we work correctly and robustly with soft mounts