python bindings do not provide retry on rate limiting errors

[Andrea Frittoli]

The openstack python bindings do not provide retries on rate limiting errors.

This is a missing feature rather than a bug.
It would be both useful for users as well as for tempest tests (see https://bugs.launchpad.net/tempest/+bug/1250809)

[Giulio Fidente (gfidente)]

I'm personally not sure how useful would this be, it seems to go against the purpose of the limits. Yet if it gets implemented maybe the clients should honor the "Retry-After" passed by the server?

[Andrea Frittoli (andrea-frittoli)]

I agree it shall honor "Retry-After" if available.

[Chmouel Boudjnah (chmouel)]

Planning to add that for python-heatclient as part of the rewrite to requests https://review.openstack.org/#/c/65704/

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/76113

[venkata anil (anil-venkata)]

Committed the changes for cinderclient for review https://review.openstack.org/#/c/76113/

[Akihiro Motoki (amotoki)]

no activity for a long time ....

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/332848

[OpenStack Infra (hudson-openstack)]

Reviewed: https://review.openstack.org/332848
Committed: https://git.openstack.org/cgit/openstack/python-cinderclient/commit/?id=f8eef18297ec2dba4abf45f8ca57c40c2380cad9
Submitter: Jenkins
Branch: master

commit f8eef18297ec2dba4abf45f8ca57c40c2380cad9
Author: Yuriy Nesenenko <email address hidden>
Date: Wed Jun 22 17:36:49 2016 +0300

    Cinder client should retry with Retry-After value

    If a request fails but the response contains a "Retry-After",
    the cinder client should wait the amount of time and then retry.
    Cinder client should report a warning to user and continue with
    retry, so that user can cancel the operation if not interested
    in retry. The value in "Retry-After" header will be in seconds
    or GMT value, client should handle both the cases.

    How many times client should retry will be controlled by user
    through "--retries" argument to cinder api example,
    $ cinder --retries 3 availability-zone-list

    If request was not sucessful within the retries, client should
    raise the exception.

    Change-Id: I99af957bfbbe3a202b148dc2fcafdd20b5d7cda0
    Partial-Bug: #1263069

[Sean McGinnis (sean-mcginnis)]

Change committed to cinderclient. It was marked as Partial, but I don't see what else needs to be done here. Please reopen if additional work is planned.

[OpenStack Infra (hudson-openstack)]

Fix proposed to branch: master
Review: https://review.openstack.org/447766

[Akihiro Motoki (amotoki)]

Currently most python bindings use keystoneauth Session as a backend, so it sounds good to start adding a support of retry-after in keystoneauth library. After that, individual python bindings can expose some parameter to control the number of retries to users.

[OpenStack Infra (hudson-openstack)]

Change abandoned by Matt Riedemann (<email address hidden>) on branch: master
Review: https://review.openstack.org/447766

[Morgan Fainberg (mdrnstm)]

This could be integrated in keystoneauth (and has been proposed as such). Marking as invalid for keystoneclient/wont fix.