Comment 9 for bug 1880777

Revision history for this message
OpenStack Infra (hudson-openstack) wrote : Fix merged to openstack-armada-app (master)

Reviewed: https://review.opendev.org/749624
Committed: https://git.openstack.org/cgit/starlingx/openstack-armada-app/commit/?id=c14bed004392f39283fab6b3039ad9333e4a3c5b
Submitter: Zuul
Branch: master

commit c14bed004392f39283fab6b3039ad9333e4a3c5b
Author: Angie Wang <email address hidden>
Date: Wed Sep 2 19:18:07 2020 -0400

    Update openstack requests to go through ingress

    The requests to Openstack services hang/fail sometimes
    due to message loss when connecting to internal service
    endpoints. This issue was observed before and fixed in
    commit https://review.opendev.org/#/c/683818/ by setting
    net.ipv4.tcp_tw_reuse to 0, however, it's still being
    seen on recent STX loads.

    It has been tested and proved that requests go through
    ingress pod do not have the issue. This commit updates
    helm charts and manifest to make all requests sending
    to openstack services go to ingress pod and then ingress
    would forward requests to the corresponding api service.

    Changes included:
    - update helm-toolkit manifest job-ks-endpoint.yaml
      to provide an ability to conditionally configure
      all types of openstack endpoints with public endpoint
      url when endpoints.identity.force_public_endpoint is
      true. Same update for keystone and keystone-api-proxy.
      With the update, for example, the admin,internal
      and public endpoints for neutron will be
      neutron.openstack.svc.cluster.local:80
    - update armada manifest to make neccessary overrides
      in openstack service configuration file to make
      communications between services go through ingress

    Change-Id: Icd646dd07d544da420a75f920bd7c3e8fc931327
    Closes-Bug: 1880777
    Signed-off-by: Angie Wang <email address hidden>