a1817f7d45
As per findings [1], the hook was found to be a burden. That is because of the lazy join on every DB lookup required to fetch the MAC for the subport. Instead of issuing a gazillion JOINs (one per each subport), this patch modifies the loop and speed it up at the expenses of some temporary memory cost: it does so by building a temporary map, looking up ports by IDs (which is bleeding fast) and reassembling the data from the temp map as the API expects it to be. Pronto! [1] https://wiki.openstack.org/wiki/Neutron_Trunk_API_Performance_and_Scaling#surprise_effect_on_filtered_port_listings Closes-bug: #bug/1648714 Change-Id: I58e39dbf1c83e1204d8d7f436395531ca096b1ab |
||
---|---|---|
.. | ||
drivers | ||
rpc | ||
seg_types | ||
__init__.py | ||
callbacks.py | ||
constants.py | ||
exceptions.py | ||
models.py | ||
plugin.py | ||
rules.py | ||
utils.py |