[Pure Storage] Replicated arrays communication failure not handled correctly

[Simon Dodsley]

It has been observed that, despite the synchronicity and uniformness being enabled, the loss of one of the arrays - be it primary or secondary, network issue (e.g. no reachability) or storage issue (e.g. pod down) - causes the Cinder driver to stop working in some or even all aspects (details depending on the particular failure that was being simulated).
This applied to both the instance-facing functionalities, such as attaching and detaching the volumes, and
general management ones, such as CRUD on volumes, or even the basic ability to restart the cinder-
volume service - the driver would fail repeatedly.
The ugly workaround is to reconfigure cinder-volume and restart it each time there was a problem with storage array, which isinflexible and prevents automatic failure recovery.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.opendev.org/c/openstack/cinder/+/855060

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/openstack/cinder/+/855060
Committed: https://opendev.org/openstack/cinder/commit/deeee2b363c4f6acf0e5cfaa6b2b04b6eb6c8362
Submitter: "Zuul (22348)"
Branch: master

commit deeee2b363c4f6acf0e5cfaa6b2b04b6eb6c8362
Author: Simon Dodsley <email address hidden>
Date: Tue May 31 14:51:51 2022 -0400

    [Pure Storage] Fix issue with loss of replicated array

    When an array on either side of a replicated configuration loses
    connectivity to Cinder this can cause the driver to stop working.

    This patch resolves that issue by validating all arrays in a
    replicated pair exist and raising a warning if that is not
    the case.

    Closes Bug: #1969784
    Co-Authored-By: Keerthivasan <email address hidden>

    Change-Id: I3256875a33a30560b834b4c8a0c6bb1b5edf69fa

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/zed
Review: https://review.opendev.org/c/openstack/cinder/+/872307

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/cinder 22.0.0.0rc1 release candidate.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.opendev.org/c/openstack/cinder/+/872307
Committed: https://opendev.org/openstack/cinder/commit/9d12d263e51c0f1b38a51571aa624862e229f203
Submitter: "Zuul (22348)"
Branch: stable/zed

commit 9d12d263e51c0f1b38a51571aa624862e229f203
Author: Simon Dodsley <email address hidden>
Date: Tue May 31 14:51:51 2022 -0400

    [Pure Storage] Fix issue with loss of replicated array

    When an array on either side of a replicated configuration loses
    connectivity to Cinder this can cause the driver to stop working.

    This patch resolves that issue by validating all arrays in a
    replicated pair exist and raising a warning if that is not
    the case.

    Closes Bug: #1969784
    Co-Authored-By: Keerthivasan <email address hidden>

    Change-Id: I3256875a33a30560b834b4c8a0c6bb1b5edf69fa
    (cherry picked from commit deeee2b363c4f6acf0e5cfaa6b2b04b6eb6c8362)

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/cinder 21.2.0 release.