NetApp+NFS in cluster mode doesn't split volumes between multiple NFS shares

[Victor Stinner]

If more than one NFS share is defined, all volumes are allocated on the same NFS share until it is not eligible anymore. Then a second NFS share is filled. The code is supposed to split Cinder volumes between the different NFS shares, not fill them one by one.

It is a bug in _find_shares(). If at least one NFS share is eligible, the "size" parameter (size of the requested volume in gigabytes) is replaced with the total size in bytes of the share. Because of that, _find_shares() cannot find more one eligible NFS share.

_find_shares() sorts eligible volumes by available size. Because of this bug, all volumes are allocated in the same NFS share, instead of being splitted on different NFS shares.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: stable/icehouse
Review: https://review.openstack.org/121467

[Eric Harney (eharney)]

This code is gone in master but applicable for icehouse.

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/121467
Committed: https://git.openstack.org/cgit/openstack/cinder/commit/?id=93de51d57b2287bfb9b39bc1351d7a612916e0a0
Submitter: Jenkins
Branch: stable/icehouse

commit 93de51d57b2287bfb9b39bc1351d7a612916e0a0
Author: Victor Stinner <email address hidden>
Date: Mon Sep 15 07:58:13 2014 +0000

    Fix NetAppDirectCmodeNfsDriver._find_shares()

    If more than one NFS share is defined, all volumes are allocated on the
    same NFS share until it is not eligible anymore. Then a second NFS share
    is filled. The code is supposed to split volumes between the different
    NFS shares, not fill them one by one.

    It is a bug in _find_shares(). If at least one NFS share is eligible,
    the "size" parameter (size of the requested volume in gigabytes) is
    replaced with the total size in bytes of the share. Because of that,
    _find_shares() cannot find more one eligible NFS share.

    _find_shares() sorts eligible volumes by available size. Because of this
    bug, all volumes are allocated in the same NFS share, instead of being
    splitted on different NFS shares.

    The _find_shares() has been removed in Juno, see:

    * commit 98aa91b0e271d6fbd9dfe43bf4152fb138a46a89
    * change Ie6f155df7bc1ae2cd5f7fa39f1b1a0ad38075988

    Change-Id: I7b72921a2a45d5dd0aa969c8c32f7d815175f3a8
    Co-author: "Florent Flament" <email address hidden>
    Closes-Bug: #1369426