"Select N+1" issue in catalog's SQL backend

[Roman Verchikov]

DESCRIPTION

SQL backend for keystone catalog has classic select n+1 issue in get_catalog() and get_v3_catalog() methods (for each endpoint get service). Even though the number of endpoints and services is usually not very high, given methods are executed on each authenticate request, which impacts performance of each openstack service regardless of keystone identity backend used.

FILES
keystone/catalog/backends/sql.py

EXPECTED BEHAVIOR
Single "select ... join" SQL statement generated

ACTUAL BEHAVIOR
N+1 select statements generated, where N is the number of endpoints.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/39386

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/39386
Committed: http://github.com/openstack/keystone/commit/b920d15b01b21831999030b0becbc1da988ee638
Submitter: Jenkins
Branch: master

commit b920d15b01b21831999030b0becbc1da988ee638
Author: Roman Verchikov <email address hidden>
Date: Wed Aug 7 14:56:24 2013 -0700

    Fix select n+1 issue in keystone catalog

    keystone.catalog.backends.sql.get_catalog() and get_v3_catalog() methods
    generate N+1 select statements for each endpoint. Use sqlalchemy's
    eager load to generate single select statement instead of generating N+1
    select statents for each endpoint.

    Given change does not modify DB schema and is runtime-only, since it's a
    one-to-many relationship.

    Change-Id: Ia72b8603fc13f01696771f6116b320364bd50f51
    Fixes: bug #1206725