Bug 1144494 - [RFE] Permit all hosts in a shared cluster to provide Local Storage just for their own use
Summary: [RFE] Permit all hosts in a shared cluster to provide Local Storage just for ...
Keywords:
Status: CLOSED DUPLICATE of bug 1134318
Alias: None
Product: ovirt-hosted-engine-setup
Classification: oVirt
Component: RFEs
Version: ---
Hardware: Unspecified
OS: Unspecified
medium
low
Target Milestone: ---
: ---
Assignee: Sandro Bonazzola
QA Contact: Pavel Stehlik
URL: http://www.ovirt.org/QA:TestCase_HA_H...
Whiteboard:
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2014-09-19 14:27 UTC by Zordrak
Modified: 2016-08-24 23:24 UTC (History)
16 users (show)

Fixed In Version:
Clone Of:
Environment:
Last Closed: 2016-08-24 23:24:57 UTC
oVirt Team: Integration
Embargoed:
ylavi: ovirt-future?
ylavi: planning_ack?
ylavi: devel_ack?
ylavi: testing_ack?


Attachments (Terms of Use)

Description Zordrak 2014-09-19 14:27:55 UTC
http://www.ovirt.org/QA:TestCase_HA_Hosted_Engine_Local_Storage

Using a HA-Hosted-Engine configuration, it is still desirable to be able to configure storage on a specific host that can only be accessed by that host.

In my specific scenario:

3x High-Capacity Hypervisors with 300GiB RAID10 15kRPM SAS each. Limited inter-host connectivity. HostedEngine installed, highly available across all three hosts using NFS served by glusterd, with gluster replicating across all three hosts and CTDB managing a shared virtual IP address. This permits the management of three hosts, from a single engine interface which may run so long as one host is on-line. One additional HA VM is run for a critical customer-facing web-service using an additional replicated NFS-accessed gluster storage domain.

Each host is then required to run additional VMs where the storage is local:

    Host A holds vdisks for VMs executed only on Host A.
    Host B holds vdisks for VMs executed only on Host B.
    Host C holds vdisks for VMs executed only on Host C. 

No live-migration is required or expected for the local VMs. Manual offline migration should be possible with the use of an NFS export domain (or scp file copy if oVirt can support manual export/import).

===

Currently this is not possible because of the hard requirement that all hosts can access all storage domains even if there is no requirement for them to be used on other hosts. To configure Local Storage each host must be in unique clusters which prevents the HA-HostedEngine configuration. Simply sharing the local storage via NFS is not possible because use of gluster-nfs and native nfsd are mutually exclusive, you cannot run both. This means the only possibilities are non-replicated gluster volumes or iSCSI exposure; both similar and unnecessarily complicated solutions. given the existing use of gluster, the gluster solution is probably better.

So, each host creates an unreplicated gluster volume and exposes the "<host>-local-data" storage domain via it's native physical IP using gluster's NFS, from where it can be used by the host on which it resides for VM backing storage.

Given the only obstacle being oVirt design architecture and not any physical or 3rd party problem, it ought to be feasible to permit each host to provide local storage only for use by itself.

Comment 1 Itamar Heim 2014-09-21 11:15:57 UTC
this is also called "SDM" - removal of the DC (SPM) as a consideration in storage domain. 
bug 1109597 also tracks the same request - allow to mix local and shared storage domains (by allowing to assign a storage domain to a host, rather than cluster/dc only)

Comment 2 Yaniv Lavi 2016-08-24 23:24:57 UTC

*** This bug has been marked as a duplicate of bug 1134318 ***


Note You need to log in before you can comment on or make changes to this bug.