| 2014-05-31 01:54:59 |
Morgan Fainberg |
bug |
|
|
added bug |
| 2014-05-31 01:55:34 |
Morgan Fainberg |
keystone: importance |
Undecided |
High |
|
| 2014-05-31 01:55:45 |
Morgan Fainberg |
keystone: status |
New |
Triaged |
|
| 2014-05-31 01:56:11 |
Morgan Fainberg |
tags |
|
icehouse-backport-potential |
|
| 2014-05-31 01:58:27 |
Morgan Fainberg |
keystone: importance |
High |
Medium |
|
| 2014-05-31 02:09:53 |
Adam Young |
description |
As discussed here: http://lists.openstack.org/pipermail/openstack-dev/2014-May/035264.htm the use of "with_lockmode('update')" can cause a number of issues when run on top of MySQL+Galera because galera does not support the 'SELECT ... FOR UPDATE' SQL call.
We currently only use with_lockmode('update') for coordinating consuming trusts (limited use trusts).
We should eliminate this and handle the coordination of consumption to ensure only the specified number of tokens can be issued from a trust. Unfortunately, this is not as straightforward as it could be, we need to handle the following deployment scenarios:
* Eventlet
* Multiple Keystone Processes (same physical server) [same issue as mod_wsgi]
* Multiple Keystone Processes (different physical servers)
The first and second ones could be handled with the lockutils (external file-based) locking decorator. The last scenario will take more thought. |
As discussed here: http://lists.openstack.org/pipermail/openstack-dev/2014-May/035264.html
the use of "with_lockmode('update')" can cause a number of issues when run on top of MySQL+Galera because galera does not support the 'SELECT ... FOR UPDATE' SQL call.
We currently only use with_lockmode('update') for coordinating consuming trusts (limited use trusts).
We should eliminate this and handle the coordination of consumption to ensure only the specified number of tokens can be issued from a trust. Unfortunately, this is not as straightforward as it could be, we need to handle the following deployment scenarios:
* Eventlet
* Multiple Keystone Processes (same physical server) [same issue as mod_wsgi]
* Multiple Keystone Processes (different physical servers)
The first and second ones could be handled with the lockutils (external file-based) locking decorator. The last scenario will take more thought. |
|
| 2014-05-31 22:34:45 |
OpenStack Infra |
keystone: status |
Triaged |
In Progress |
|
| 2014-05-31 22:34:45 |
OpenStack Infra |
keystone: assignee |
|
Morgan Fainberg (mdrnstm) |
|
| 2014-07-22 22:09:51 |
Morgan Fainberg |
keystone: milestone |
|
juno-3 |
|
| 2014-08-06 19:11:01 |
OpenStack Infra |
keystone: status |
In Progress |
Fix Committed |
|
| 2014-08-25 21:21:19 |
Morgan Fainberg |
nominated for series |
|
keystone/icehouse |
|
| 2014-08-25 21:21:19 |
Morgan Fainberg |
bug task added |
|
keystone/icehouse |
|
| 2014-08-25 21:21:27 |
Morgan Fainberg |
keystone/icehouse: status |
New |
In Progress |
|
| 2014-08-25 21:21:30 |
Morgan Fainberg |
keystone/icehouse: importance |
Undecided |
Medium |
|
| 2014-08-25 21:21:33 |
Morgan Fainberg |
keystone/icehouse: assignee |
|
Morgan Fainberg (mdrnstm) |
|
| 2014-09-04 14:32:12 |
Thierry Carrez |
keystone: status |
Fix Committed |
Fix Released |
|
| 2014-09-25 23:35:54 |
OpenStack Infra |
keystone/icehouse: status |
In Progress |
Fix Committed |
|
| 2014-09-29 19:48:41 |
Adam Gandelman |
keystone/icehouse: milestone |
|
2014.1.3 |
|
| 2014-10-02 23:43:34 |
Adam Gandelman |
keystone/icehouse: status |
Fix Committed |
Fix Released |
|
| 2014-10-16 08:19:51 |
Thierry Carrez |
keystone: milestone |
juno-3 |
2014.2 |
|
