On 09/20/2012 05:37 PM, Roland Kool wrote:
Hi all,
Hopefully this is the correct place to ask this, if not, my apologies.
On a CS 3.0.2 setup using KVM on RHEL6 and NFS as primary storage, we noted
some strange behaviour when it comes down to the NFS mount options for the
primary storage.
When a hypervisor is booted, the primary storage is mounted using the default
mount options as configured in /etc/nfsmount.conf. However, when we shutdown
the agent, manually unmount the primary storage and start the cloud agent
again, the primary storage is mounted using some hardcoded values (in our case
the NFS mount becomes a soft mount, in stead of hard and the timeo and retrans
settings change).
After some digging, we found that the script
/usr/lib64/cloud/agent/scripts/vm/hypervisor/kvm/kvmheartbeat.sh is responsible
for that (the mountoptions are hardcoded there).
My question is: what's the reasoning behind this? Is this a bug or leftover? Or
is this by design and if so, why are these settings now always enforced?
This is a leftover and legacy. You're right that you should be able to
specify your own NFS mount options.
As the number of storage pool options in CloudStack is expanding we are
looking at a refactor of the storage code, one of the things is that
these kind of options can be submitted by the admin.
So yes, it's on our radar, we haven't been able to look at this yet though.
Wido
Thanks in advance for any insights and your time.
Best regards,
Roland Kool | Sr. Systems Engineer | TomTom TechOps |
