Comment 11 for bug 1810563

Reviewed: https://review.openstack.org/633143
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=f8a192e22e39bfb64e506388f5d5c07fc30d7753
Submitter: Zuul
Branch: stable/rocky

commit f8a192e22e39bfb64e506388f5d5c07fc30d7753
Author: Doug Wiegley <email address hidden>
Date: Fri Jan 4 14:55:29 2019 -0700

    Fix performance regression adding rules to security groups

    Sometime between liberty and pike, adding rules to SG's got
    slow, and slower with every rule. Streamline the rule create path,
    and get close to the old performance back.

    Two performance fixes:
    1. Get rid of an n^2 duplicate check, using a hash table instead,
    on bulk creates. This is more memory intensive than the previous loop,
    but usable far past where the other becomes too slow to be useful.
    2. Use an object existence check in a few places where we do not
    want to load all of the child rules.

    Also squashed in:
    Restore tenant_id check on security group rule adds to previous semantic

    We switched from swapping the tenant_id in the context to explicitly
    checking the db column. Switch back, and a test that checks for
    not breaking this rather odd behavior. At least, until we decide
    to fix it as a bug.

    Co-Authored-By: William Hager <email address hidden>
    Change-Id: I34e41a128f28211f2e7ab814a2611ce22620fcf3
    Closes-bug: 1810563
    (cherry picked from commit 2eb31f84c9a6c9fc6340819f756a7a82cbf395f3)
    (squashed patch from commit bd4c291cdfb5a75976b1f846f6d7ca88d2d154e9)