Hide Forgot
This release adds a Technology Preview of the ability to attach a volume to multiple hosts or servers simultaneously in both cinder and nova with read/write (RW) mode when this feature is supported by the back end driver. This feature addresses the clustered application workloads use case that typically requires active/active or active/standby scenarios.
Cloned from launchpad blueprint https://blueprints.launchpad.net/cinder/+spec/multi-attach-volume. Description: Currently, the existing block storage (cinder) drivers allow for only a single volume to be attached to a single VM instance via iSCSI or Fibre Channel. To support a cluster, a single volume would need to be exported to multiple host. In this case vCenter will be used as an example, but any Hypervisor that supports clusters could require this functionality. The Cinder etherpad https://etherpad.openstack.org/summit-havana-cinder-multi-attach-and-ro-volumes was present at the Havana Design summit resulted in the majority of the work to be completed in nova project. This work includes the looping mechanism that's responsible for calling the cinder drivers for each host in the cluster. The Cinder work is not yet determined but would require minor work in tracking that a volume has multiple host attached to it. Possibly changing the database entries for what volume is attached to a single host to a list of host. More work may be required once test have been ran to verify the data flow between multiple attach points. This blueprint in for the limited cinder work that would be required from the following Nova blueprint: https://blueprints.launchpad.net/nova/+spec/fc-support-for-vcenter-driver Specification URL (additional information): None
*** Bug 1033178 has been marked as a duplicate of this bug. ***
Feature specs: https://review.openstack.org/#/c/154135/8/specs/kilo/multi-attach-volume.rst
This bug was accidentally moved from POST to MODIFIED via an error in automation, please see mmccune@redhat.com with any questions
cinderclient 1.3.0 python-cinderclient-1.8.1-0.20160607191316.c955397.el7ost (which is in OSP10) is necessary but insufficient to address this bug. Per discussion, adding needinfo on eharney to clarify details of remaining work. Moving to RHOS 11, as this remaining work is by his estimation very unlikely to occur in 10.
Current multi-attach blueprint for Nova: https://blueprints.launchpad.net/nova/+spec/multi-attach-volume Nova Spec: https://review.openstack.org/#/c/304681/
(In reply to Eric Harney from comment #15) > Current multi-attach blueprint for Nova: > https://blueprints.launchpad.net/nova/+spec/multi-attach-volume > > Nova Spec: https://review.openstack.org/#/c/304681/ I believe the downstream tracker for these is https://bugzilla.redhat.com/show_bug.cgi?id=1033185
*** Bug 1230807 has been marked as a duplicate of this bug. ***
Upstream specs: https://blueprints.launchpad.net/nova/+spec/multi-attach-volume https://blueprints.launchpad.net/cinder/+spec/multi-attach-v3-attach
Benny hit Nova issues bz1623365 while testing automation for this RFE.
Verified according to test plan RPM package version : python-cinder-13.0.1-0.20181013185427.31ff628.el7ost.noarch python2-cinderclient-4.0.1-0.20180809133302.460229c.el7ost.noarch puppet-cinder-13.3.1-0.20181013114719.25b1ba3.el7ost.noarch openstack-cinder-13.0.1-0.20181013185427.31ff628.el7ost.noarch
Since the problem described in this bug report should be resolved in a recent advisory, it has been closed with a resolution of ERRATA. For information on the advisory, and where to find the updated files, follow the link below. If the solution does not work for you, open a new bug report. https://access.redhat.com/errata/RHEA-2019:0045
This enhancement lets you attach a Block Storage volume to multiple hosts. You can attach a volume to multiple hosts or servers simultaneously in both the Block Storage (cinder) and Compute (nova) services with read/write (RW) mode. The feature addresses the clustered application workloads use case, which typically requires active/active or active/standby scenarios. The feature is intended for use with basic volume operations. Known limitations include encrypted volumes and migration. Availability depends on back end support.