sqlalchemy-driver: Whole service stops working if DB connection is not working

[Thomas Bechtold]

When setting up osprofiler with the sqlalchemy driver, the whole service (eg. keystone, nova, ...) can be killed when eg. the connection string for the database is not correct.

This is not good. The service should continue to work even if there is a problem with initializing/using osprofiler.

[Thomas Bechtold (toabctl)]

See https://review.openstack.org/#/c/642407/

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/642407
Committed: https://git.openstack.org/cgit/openstack/osprofiler/commit/?id=add4794120fa7cfbc1be1b6b804298a95da1b5ff
Submitter: Zuul
Branch: master

commit add4794120fa7cfbc1be1b6b804298a95da1b5ff
Author: Thomas Bechtold <email address hidden>
Date: Mon Mar 11 11:20:19 2019 +0100

    Don't fail if sqlalchemy driver fails to initialize

    Given that the driver is initialized during the service
    startup (eg. like Keystone, Nova, ...) a osprofiler driver failure has
    a huge impact on the service (when the driver fails, the whole service
    is not usable).
    We want to avoid that and just log error/exceptions but keep the
    services running.

    Change-Id: I5688f10364884a74b7eb44c0c8bda15730ccd424
    Closes-Bug: 1819433

[OpenStack Infra (hudson-openstack)]

This issue was fixed in the openstack/osprofiler 2.7.0 release.