MySQL deadlock creating share networks concurrently

[Valeriy Ponomaryov]

If we create lots of share networks at once we can get "deadlock" using MySQL DB back end.

Here are logs [1] and [2]

[1] http://logs.openstack.org/11/203911/9/check/gate-rally-dsvm-manila-multibackend/9263110/logs/screen-m-api.txt.gz?level=TRACE

[2] http://paste.openstack.org/show/412508/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/211484

[Valeriy Ponomaryov (vponomaryov)]

One more deadlock hit - http://logs.openstack.org/29/188029/12/experimental/gate-rally-dsvm-manila-multibackend-no-ss/688360d/logs/screen-m-api.txt.gz?level=TRACE

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/211484
Committed: https://git.openstack.org/cgit/openstack/manila/commit/?id=ab05b5c4651f400b7aaab803be4ace8c6e0bc114
Submitter: Jenkins
Branch: master

commit ab05b5c4651f400b7aaab803be4ace8c6e0bc114
Author: Valeriy Ponomaryov <email address hidden>
Date: Tue Aug 11 12:53:56 2015 +0300

    Add retries for deadlock-vulnerable DB methods

    It is very likely to get 'deadlock' updating common records such as
    quotas, usages, reservations and metadata. All of these things are
    API-dependent. So, add wrappers, that retries request on 'deadlock'
    exceptions.
    Also, set database config opt "max_pool_size" to '40' instead of having
    default value - '5'. It allows to keep upto 40 SQL connection alive at
    once.

    Change-Id: Ia6ec09d4617191c51ad95dd6d9f053141ec1c4c4
    Closes-Bug: #1483604