api and rpc worker defaults are problematic

[Doug Wiegley]

We default the number of api workers to the number of cores. At approximately 2GB per neutron-server, sometimes that's more RAM than is available, and the OOM killer comes out.

We default the number of rpc workers to 1, which seems to fall behind on all but the smallest deployments.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/636363

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/636363
Committed: https://git.openstack.org/cgit/openstack/neutron/commit/?id=7e09b25b964dde82caf5f5159d25810b6f8ebd3c
Submitter: Zuul
Branch: master

commit 7e09b25b964dde82caf5f5159d25810b6f8ebd3c
Author: Doug Wiegley <email address hidden>
Date: Tue Feb 12 08:47:19 2019 -0700

    Modify api and rpc default number of workers

    - Limit number of api workers to roughly using half of system
      RAM. Spawning a bunch, just to have the OOM killer nuke them
      regularly is not useful.
    - Bump the rpc_workers default to half of the api_workers.
      A default of 1 falls behind on any reasonably sized node.

    Change-Id: I8b84a359f83133014b3d4414aafc10e6b7c6a876
    Closes-bug: #1815629

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/neutron 14.0.0.0b2 development milestone.